Bump version to 5.7.3

Summary:
fixed in #2799
Closes https://github.com/facebook/rocksdb/pull/2805

Differential Revision: D5734324

Pulled By: ajkr

fbshipit-source-id: a285d4e84bf1018dc2257fd6c3e7c075a7243263

CMakeLists.txt

@@ -380,6 +380,7 @@ set(SOURCES
         db/write_thread.cc
         env/env.cc
         env/env_chroot.cc
+        env/env_encryption.cc
         env/env_hdfs.cc
         env/mock_env.cc
         memtable/alloc_tracker.cc
@@ -482,6 +483,7 @@ set(SOURCES
         utilities/blob_db/blob_log_reader.cc
         utilities/blob_db/blob_log_writer.cc
         utilities/blob_db/blob_log_format.cc
+        utilities/blob_db/ttl_extractor.cc
         utilities/checkpoint/checkpoint_impl.cc
         utilities/col_buf_decoder.cc
         utilities/col_buf_encoder.cc

HISTORY.md

@@ -1,11 +1,32 @@
 # Rocksdb Change Log
-## 5.6.2 (08/12/2017)
+## 5.7.3 (08/29/2017)
+### Bug Fixes
+* Fix transient reappearance of keys covered by range deletions when memtable prefix bloom filter is enabled.
+* Fix potentially wrong file smallest key when range deletions separated by snapshot are written together.
+
+## 5.7.2 (08/15/2017)
+### Bug Fixes
+* Fix incorrect dropping of deletions issue with FIFO compaction.
+* Fix LITE build compiler error with missing abort().
+
+## 5.7.1 (08/13/2017)
 ### Bug Fixes
 * Fix incorrect dropping of deletions during intra-L0 compaction.
 
-## 5.6.1 (07/25/2017)
+## 5.7.0 (07/13/2017)
+### Public API Change
+* DB property "rocksdb.sstables" now prints keys in hex form.
+
+### New Features
+* Measure estimated number of reads per file. The information can be accessed through DB::GetColumnFamilyMetaData or "rocksdb.sstables" DB property.
+* RateLimiter support for throttling background reads, or throttling the sum of background reads and writes. This can give more predictable I/O usage when compaction reads more data than it writes, e.g., due to lots of deletions.
+* [Experimental] FIFO compaction with TTL support. It can be enabled by setting CompactionOptionsFIFO.ttl > 0.
+* Introduce `EventListener::OnBackgroundError()` callback. Users can implement it to be notified of errors causing the DB to enter read-only mode, and optionally override them.
+* Partitioned Index/Filters exiting the experimental mode. To enable partitioned indexes set index_type to kTwoLevelIndexSearch and to further enable partitioned filters set partition_filters to true. To configure the partition size set metadata_block_size.
+
+
 ### Bug Fixes
-* Fix lite build.
+* Fix discarding empty compaction output files when `DeleteRange()` is used together with subcompactions.
 
 ## 5.6.0 (06/06/2017)
 ### Public API Change

INSTALL.md

@@ -32,6 +32,7 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
 * You can link RocksDB with following compression libraries:
   - [zlib](http://www.zlib.net/) - a library for data compression.
   - [bzip2](http://www.bzip.org/) - a library for data compression.
+  - [lz4](https://github.com/lz4/lz4) - a library for extremely fast data compression.
   - [snappy](http://google.github.io/snappy/) - a library for fast
       data compression.
   - [zstandard](http://www.zstd.net) - Fast real-time compression
@@ -53,6 +54,7 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
       `sudo apt-get install libsnappy-dev`.
     * Install zlib. Try: `sudo apt-get install zlib1g-dev`.
     * Install bzip2: `sudo apt-get install libbz2-dev`.
+    * Install lz4: `sudo apt-get install liblz4-dev`.
     * Install zstandard: `sudo apt-get install libzstd-dev`.
 
 * **Linux - CentOS / RHEL**
@@ -81,6 +83,10 @@ to build a portable binary, add `PORTABLE=1` before your make commands, like thi
 
               sudo yum install bzip2 bzip2-devel
 
+    * Install lz4:
+
+              sudo yum install lz4-devel
+
     * Install ASAN (optional for debugging):
 
               sudo yum install libasan

Makefile

@@ -323,6 +323,7 @@ EXPOBJECTS = $(EXP_LIB_SOURCES:.cc=.o) $(LIBOBJECTS) $(TESTUTIL)
 
 TESTS = \
 	db_basic_test \
+	db_encryption_test \
 	db_test2 \
 	external_sst_file_basic_test \
 	auto_roll_logger_test \
@@ -1035,6 +1036,9 @@ slice_transform_test: util/slice_transform_test.o $(LIBOBJECTS) $(TESTHARNESS)
 db_basic_test: db/db_basic_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 
+db_encryption_test: db/db_encryption_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
+	$(AM_LINK)
+
 db_test: db/db_test.o db/db_test_util.o $(LIBOBJECTS) $(TESTHARNESS)
 	$(AM_LINK)
 

TARGETS

@@ -110,6 +110,7 @@ cpp_library(
       "db/write_thread.cc",
       "env/env.cc",
       "env/env_chroot.cc",
+      "env/env_encryption.cc",
       "env/env_hdfs.cc",
       "env/env_posix.cc",
       "env/io_posix.cc",
@@ -211,6 +212,7 @@ cpp_library(
       "utilities/blob_db/blob_log_reader.cc",
       "utilities/blob_db/blob_log_writer.cc",
       "utilities/blob_db/blob_log_format.cc",
+      "utilities/blob_db/ttl_extractor.cc",
       "utilities/checkpoint/checkpoint_impl.cc",
       "utilities/compaction_filters/remove_emptyvalue_compactionfilter.cc",
       "utilities/convenience/info_log_finder.cc",

USERS.md

@@ -76,7 +76,7 @@ quasardb uses a heavily tuned RocksDB as its persistence layer.
 [Dgraph](https://github.com/dgraph-io/dgraph) is an open-source, scalable, distributed, low latency, high throughput Graph database .They use RocksDB to store state locally on a machine.
 
 ## Uber
-[Uber](http://eng.uber.com/cherami/) uses RocksDB as a durable and and scalable task queue.
+[Uber](http://eng.uber.com/cherami/) uses RocksDB as a durable and scalable task queue.
 
 ## 360 Pika
 [360](http://www.360.cn/) [Pika](https://github.com/Qihoo360/pika) is a nosql compatible with redis. With the huge amount of data stored, redis may suffer for a capacity bottleneck, and pika was born for solving it. It has widely been widely used in many company

arcanist_util/INTERNAL_ONLY_DIR

@@ -1,2 +0,0 @@
-arcanist_util are only used internaly, If you want to change it please check
-<internal_rocksdb_repo>/arcanist_util

arcanist_util/__phutil_library_init__.php

@@ -0,0 +1,3 @@
+<?php
+
+phutil_register_library('arcanist_util', __FILE__);

arcanist_util/__phutil_library_map__.php

@@ -0,0 +1,71 @@
+<?php
+
+/**
+ * This file is automatically generated. Use 'arc liberate' to rebuild it.
+ * @generated
+ * @phutil-library-version 2
+ */
+
+if (class_exists('ArcanistWorkflow')) {
+  phutil_register_library_map(array(
+    '__library_version__' => 2,
+    'class' =>
+    array(
+      'ArcanistCpplintLinter' => 'cpp_linter/ArcanistCpplintLinter.php',
+      'BaseDirectoryScopedFormatLinter' => 'cpp_linter/BaseDirectoryScopedFormatLinter.php',
+      'FacebookArcanistConfiguration' => 'config/FacebookArcanistConfiguration.php',
+      'FacebookFbcodeLintEngine' => 'lint_engine/FacebookFbcodeLintEngine.php',
+      'FacebookFbcodeUnitTestEngine' => 'unit_engine/FacebookFbcodeUnitTestEngine.php',
+      'FacebookHowtoevenLintEngine' => 'lint_engine/FacebookHowtoevenLintEngine.php',
+      'FacebookHowtoevenLinter' => 'cpp_linter/FacebookHowtoevenLinter.php',
+      'FbcodeClangFormatLinter' => 'cpp_linter/FbcodeClangFormatLinter.php',
+      'FbcodeCppLinter' => 'cpp_linter/FbcodeCppLinter.php',
+    ),
+    'function' =>
+    array(
+    ),
+    'xmap' =>
+    array(
+      'ArcanistCpplintLinter' => 'ArcanistLinter',
+      'BaseDirectoryScopedFormatLinter' => 'ArcanistLinter',
+      'FacebookArcanistConfiguration' => 'ArcanistConfiguration',
+      'FacebookFbcodeLintEngine' => 'ArcanistLintEngine',
+      'FacebookFbcodeUnitTestEngine' => 'ArcanistBaseUnitTestEngine',
+      'FacebookHowtoevenLintEngine' => 'ArcanistLintEngine',
+      'FacebookHowtoevenLinter' => 'ArcanistLinter',
+      'FbcodeClangFormatLinter' => 'BaseDirectoryScopedFormatLinter',
+      'FbcodeCppLinter' => 'ArcanistLinter',
+    ),
+  ));
+} else {
+  phutil_register_library_map(array(
+    '__library_version__' => 2,
+    'class' =>
+    array(
+      'ArcanistCpplintLinter' => 'cpp_linter/ArcanistCpplintLinter.php',
+      'BaseDirectoryScopedFormatLinter' => 'cpp_linter/BaseDirectoryScopedFormatLinter.php',
+      'FacebookArcanistConfiguration' => 'config/FacebookOldArcanistConfiguration.php',
+      'FacebookFbcodeLintEngine' => 'lint_engine/FacebookFbcodeLintEngine.php',
+      'FacebookFbcodeUnitTestEngine' => 'unit_engine/FacebookOldFbcodeUnitTestEngine.php',
+      'FacebookHowtoevenLintEngine' => 'lint_engine/FacebookHowtoevenLintEngine.php',
+      'FacebookHowtoevenLinter' => 'cpp_linter/FacebookHowtoevenLinter.php',
+      'FbcodeClangFormatLinter' => 'cpp_linter/FbcodeClangFormatLinter.php',
+      'FbcodeCppLinter' => 'cpp_linter/FbcodeCppLinter.php',
+    ),
+    'function' =>
+    array(
+    ),
+    'xmap' =>
+    array(
+      'ArcanistCpplintLinter' => 'ArcanistLinter',
+      'BaseDirectoryScopedFormatLinter' => 'ArcanistLinter',
+      'FacebookArcanistConfiguration' => 'ArcanistConfiguration',
+      'FacebookFbcodeLintEngine' => 'ArcanistLintEngine',
+      'FacebookFbcodeUnitTestEngine' => 'ArcanistBaseUnitTestEngine',
+      'FacebookHowtoevenLintEngine' => 'ArcanistLintEngine',
+      'FacebookHowtoevenLinter' => 'ArcanistLinter',
+      'FbcodeClangFormatLinter' => 'BaseDirectoryScopedFormatLinter',
+      'FbcodeCppLinter' => 'ArcanistLinter',
+    ),
+  ));
+}

arcanist_util/config/FacebookArcanistConfiguration.php

@@ -0,0 +1,43 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+require('RocksDBCommonHelper.php');
+
+define("DIFF_COMMAND", "diff");
+
+class FacebookArcanistConfiguration extends ArcanistConfiguration {
+  public function getCustomArgumentsForCommand($command) {
+    if ($command == "land") {
+      return array(
+          'async' => array('help' => 'Just to make tools happy'));
+    }
+    return array();
+  }
+
+  public function didRunWorkflow($command,
+                                 ArcanistWorkflow $workflow,
+                                 $error_code) {
+    // Default options don't terminate on failure, but that's what we want. In
+    // the current case we use assertions intentionally as "terminate on failure
+    // invariants".
+    assert_options(ASSERT_BAIL, true);
+
+    assert($workflow);
+    assert(strlen($command) > 0);
+
+    if ($command == DIFF_COMMAND && !$workflow->isRawDiffSource()) {
+      $diffID = $workflow->getDiffId();
+
+      // When submitting a diff this code path gets executed multiple times in
+      // a row. We only care about the case when ID for the diff is provided
+      // because that's what we need to apply the diff and trigger the tests.
+      if (strlen($diffID) > 0) {
+        assert(is_numeric($diffID));
+        startTestsInSandcastle(true /* $applyDiff */, $workflow, $diffID);
+      }
+    }
+  }
+}

arcanist_util/config/FacebookOldArcanistConfiguration.php

@@ -0,0 +1,43 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+require('RocksDBCommonHelper.php');
+
+define("DIFF_COMMAND", "diff");
+
+class FacebookArcanistConfiguration extends ArcanistConfiguration {
+  public function getCustomArgumentsForCommand($command) {
+    if ($command == "land") {
+      return array(
+          'async' => array('help' => 'Just to make tools happy'));
+    }
+    return array();
+  }
+
+  public function didRunWorkflow($command,
+                                 ArcanistBaseWorkflow $workflow,
+                                 $error_code) {
+    // Default options don't terminate on failure, but that's what we want. In
+    // the current case we use assertions intentionally as "terminate on failure
+    // invariants".
+    assert_options(ASSERT_BAIL, true);
+
+    assert($workflow);
+    assert(strlen($command) > 0);
+
+    if ($command == DIFF_COMMAND && !$workflow->isRawDiffSource()) {
+      $diffID = $workflow->getDiffId();
+
+      // When submitting a diff this code path gets executed multiple times in
+      // a row. We only care about the case when ID for the diff is provided
+      // because that's what we need to apply the diff and trigger the tests.
+      if (strlen($diffID) > 0) {
+        assert(is_numeric($diffID));
+        startTestsInSandcastle(true /* $applyDiff */, $workflow, $diffID);
+      }
+    }
+  }
+}

arcanist_util/config/RocksDBCommonHelper.php

@@ -0,0 +1,355 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+// Name of the environment variables which need to be set by the entity which
+// triggers continuous runs so that code at the end of the file gets executed
+// and Sandcastle run starts.
+define("ENV_POST_RECEIVE_HOOK", "POST_RECEIVE_HOOK");
+define("ENV_HTTPS_APP_VALUE", "HTTPS_APP_VALUE");
+define("ENV_HTTPS_TOKEN_VALUE", "HTTPS_TOKEN_VALUE");
+
+define("PRIMARY_TOKEN_FILE", '/home/krad/.sandcastle');
+define("CONT_RUN_ALIAS", "leveldb");
+
+//////////////////////////////////////////////////////////////////////
+/*  Run tests in sandcastle */
+function postURL($diffID, $url) {
+  assert(strlen($diffID) > 0);
+  assert(is_numeric($diffID));
+  assert(strlen($url) > 0);
+
+  $cmd = 'echo \'{"diff_id": ' . $diffID . ', '
+         . '"name":"click here for sandcastle tests for D' . $diffID . '", '
+         . '"link":"' . $url . '"}\' | '
+         . 'arc call-conduit '
+         . 'differential.updateunitresults';
+  shell_exec($cmd);
+}
+
+function buildUpdateTestStatusCmd($diffID, $test, $status) {
+  assert(strlen($diffID) > 0);
+  assert(is_numeric($diffID));
+  assert(strlen($test) > 0);
+  assert(strlen($status) > 0);
+
+  $cmd = 'echo \'{"diff_id": ' . $diffID . ', '
+         . '"name":"' . $test . '", '
+         . '"result":"' . $status . '"}\' | '
+         . 'arc call-conduit '
+         . 'differential.updateunitresults';
+  return $cmd;
+}
+
+function updateTestStatus($diffID, $test) {
+  assert(strlen($diffID) > 0);
+  assert(is_numeric($diffID));
+  assert(strlen($test) > 0);
+
+  shell_exec(buildUpdateTestStatusCmd($diffID, $test, "waiting"));
+}
+
+function getSteps($applyDiff, $diffID, $username, $test) {
+  assert(strlen($username) > 0);
+  assert(strlen($test) > 0);
+
+  if ($applyDiff) {
+    assert(strlen($diffID) > 0);
+    assert(is_numeric($diffID));
+
+    $arcrc_content = (PHP_OS == "Darwin" ?
+        exec("cat ~/.arcrc | gzip -f | base64") :
+            exec("cat ~/.arcrc | gzip -f | base64 -w0"));
+    assert(strlen($arcrc_content) > 0);
+
+    // Sandcastle machines don't have arc setup. We copy the user certificate
+    // and authenticate using that in Sandcastle.
+    $setup = array(
+      "name" => "Setup arcrc",
+      "shell" => "echo " . $arcrc_content . " | base64 --decode"
+                 . " | gzip -d > ~/.arcrc",
+      "user" => "root"
+    );
+
+    // arc demands certain permission on its config.
+    // also fix the sticky bit issue in sandcastle
+    $fix_permission = array(
+      "name" => "Fix environment",
+      "shell" => "chmod 600 ~/.arcrc && chmod +t /dev/shm",
+      "user" => "root"
+    );
+
+    // Construct the steps in the order of execution.
+    $steps[] = $setup;
+    $steps[] = $fix_permission;
+  }
+
+  // fbcode is a sub-repo. We cannot patch until we add it to ignore otherwise
+  // Git thinks it is an uncommited change.
+  $fix_git_ignore = array(
+    "name" => "Fix git ignore",
+    "shell" => "echo fbcode >> .git/info/exclude",
+    "user" => "root"
+  );
+
+  $steps[] = $fix_git_ignore;
+
+  // This will be the command used to execute particular type of tests.
+  $cmd = "";
+
+  if ($applyDiff) {
+    // Patch the code (keep your fingures crossed).
+    $patch = array(
+      "name" => "Patch " . $diffID,
+      "shell" => "arc --arcrc-file ~/.arcrc "
+                  . "patch --nocommit --diff " . $diffID,
+      "user" => "root"
+    );
+
+    $steps[] = $patch;
+
+    updateTestStatus($diffID, $test);
+    $cmd = buildUpdateTestStatusCmd($diffID, $test, "running") . "; ";
+  }
+
+  // Run the actual command.
+  $cmd = $cmd . "J=$(nproc) ./build_tools/precommit_checker.py " . $test
+           . "; exit_code=$?; ";
+
+  if ($applyDiff) {
+    $cmd = $cmd . "([[ \$exit_code -eq 0 ]] &&"
+                . buildUpdateTestStatusCmd($diffID, $test, "pass") . ")"
+                . "||" . buildUpdateTestStatusCmd($diffID, $test, "fail")
+                . "; ";
+  }
+
+  // shell command to sort the tests based on exit code and print
+  // the output of the log files.
+  $cat_sorted_logs = "
+    while read code log_file;
+      do echo \"################ cat \$log_file [exit_code : \$code] ################\";
+      cat \$log_file;
+    done < <(tail -n +2 LOG | sort -k7,7n -k4,4gr | awk '{print \$7,\$NF}')";
+
+  // Shell command to cat all log files
+  $cat_all_logs = "for f in `ls t/!(run-*)`; do echo \$f;cat \$f; done";
+
+  // If LOG file exist use it to cat log files sorted by exit code, otherwise
+  // cat everything
+  $logs_cmd = "if [ -f LOG ]; then {$cat_sorted_logs}; else {$cat_all_logs}; fi";
+
+  $cmd = $cmd . " cat /tmp/precommit-check.log"
+              . "; shopt -s extglob; {$logs_cmd}"
+              . "; shopt -u extglob; [[ \$exit_code -eq 0 ]]";
+  assert(strlen($cmd) > 0);
+
+  $run_test = array(
+    "name" => "Run " . $test,
+    "shell" => $cmd,
+    "user" => "root",
+    "parser" => "python build_tools/error_filter.py " . $test,
+  );
+
+  $steps[] = $run_test;
+
+  if ($applyDiff) {
+    // Clean up the user arc config we are using.
+    $cleanup = array(
+      "name" => "Arc cleanup",
+      "shell" => "rm -f ~/.arcrc",
+      "user" => "root"
+    );
+
+    $steps[] = $cleanup;
+  }
+
+  assert(count($steps) > 0);
+  return $steps;
+}
+
+function getSandcastleConfig() {
+  $sandcastle_config = array();
+
+  $cwd = getcwd();
+  $cwd_token_file = "{$cwd}/.sandcastle";
+  // This is a case when we're executed from a continuous run. Fetch the values
+  // from the environment.
+  if (getenv(ENV_POST_RECEIVE_HOOK)) {
+    $sandcastle_config[0] = getenv(ENV_HTTPS_APP_VALUE);
+    $sandcastle_config[1] = getenv(ENV_HTTPS_TOKEN_VALUE);
+  } else {
+    // This is a typical `[p]arc diff` case. Fetch the values from the specific
+    // configuration files.
+    for ($i = 0; $i < 50; $i++) {
+      if (file_exists(PRIMARY_TOKEN_FILE) ||
+          file_exists($cwd_token_file)) {
+        break;
+      }
+      // If we failed to fetch the tokens, sleep for 0.2 second and try again
+      usleep(200000);
+    }
+    assert(file_exists(PRIMARY_TOKEN_FILE) ||
+           file_exists($cwd_token_file));
+
+    // Try the primary location first, followed by a secondary.
+    if (file_exists(PRIMARY_TOKEN_FILE)) {
+      $cmd = 'cat ' . PRIMARY_TOKEN_FILE;
+    } else {
+      $cmd = 'cat ' . $cwd_token_file;
+    }
+
+    assert(strlen($cmd) > 0);
+    $sandcastle_config = explode(':', rtrim(shell_exec($cmd)));
+  }
+
+  // In this case be very explicit about the implications.
+  if (count($sandcastle_config) != 2) {
+    echo "Sandcastle configuration files don't contain valid information " .
+         "or the necessary environment variables aren't defined. Unable " .
+         "to validate the code changes.";
+    exit(1);
+  }
+
+  assert(strlen($sandcastle_config[0]) > 0);
+  assert(strlen($sandcastle_config[1]) > 0);
+  assert(count($sandcastle_config) > 0);
+
+  return $sandcastle_config;
+}
+
+// This function can be called either from `[p]arc diff` command or during
+// the Git post-receive hook.
+ function startTestsInSandcastle($applyDiff, $workflow, $diffID) {
+  // Default options don't terminate on failure, but that's what we want. In
+  // the current case we use assertions intentionally as "terminate on failure
+  // invariants".
+  assert_options(ASSERT_BAIL, true);
+
+  // In case of a diff we'll send notificatios to the author. Else it'll go to
+  // the entire team because failures indicate that build quality has regressed.
+  $username = $applyDiff ? exec("whoami") : CONT_RUN_ALIAS;
+  assert(strlen($username) > 0);
+
+  if ($applyDiff) {
+    assert($workflow);
+    assert(strlen($diffID) > 0);
+    assert(is_numeric($diffID));
+  }
+
+  // List of tests we want to run in Sandcastle.
+  $tests = array("unit", "unit_non_shm", "unit_481", "clang_unit", "tsan",
+                 "asan", "lite_test", "valgrind", "release", "release_481",
+                 "clang_release", "punit", "clang_analyze", "code_cov",
+                 "java_build", "no_compression", "unity", "ubsan");
+
+  $send_email_template = array(
+    'type' => 'email',
+    'triggers' => array('fail'),
+    'emails' => array($username . '@fb.com'),
+  );
+
+  // Construct a job definition for each test and add it to the master plan.
+  foreach ($tests as $test) {
+    $stepName = "RocksDB diff " . $diffID . " test " . $test;
+
+    if (!$applyDiff) {
+      $stepName = "RocksDB continuous integration test " . $test;
+    }
+
+    $arg[] = array(
+      "name" => $stepName,
+      "report" => array($send_email_template),
+      "steps" => getSteps($applyDiff, $diffID, $username, $test)
+    );
+  }
+
+  // We cannot submit the parallel execution master plan to Sandcastle and
+  // need supply the job plan as a determinator. So we construct a small job
+  // that will spit out the master job plan which Sandcastle will parse and
+  // execute. Why compress the job definitions? Otherwise we run over the max
+  // string size.
+  $cmd = "echo " . base64_encode(json_encode($arg))
+         . (PHP_OS == "Darwin" ?
+             " | gzip -f | base64" :
+                 " | gzip -f | base64 -w0");
+  assert(strlen($cmd) > 0);
+
+  $arg_encoded = shell_exec($cmd);
+  assert(strlen($arg_encoded) > 0);
+
+  $runName = "Run diff " . $diffID . "for user " . $username;
+
+  if (!$applyDiff) {
+    $runName = "RocksDB continuous integration build and test run";
+  }
+
+  $command = array(
+    "name" => $runName,
+    "steps" => array()
+  );
+
+  $command["steps"][] = array(
+    "name" => "Generate determinator",
+    "shell" => "echo " . $arg_encoded . " | base64 --decode | gzip -d"
+               . " | base64 --decode",
+    "determinator" => true,
+    "user" => "root"
+  );
+
+  // Submit to Sandcastle.
+  $url = 'https://interngraph.intern.facebook.com/sandcastle/create';
+
+  $job = array(
+    'command' => 'SandcastleUniversalCommand',
+    'args' => $command,
+    'capabilities' => array(
+      'vcs' => 'rocksdb-int-git',
+      'type' => 'lego',
+    ),
+    'hash' => 'origin/master',
+    'user' => $username,
+    'alias' => 'rocksdb-precommit',
+    'tags' => array('rocksdb'),
+    'description' => 'Rocksdb precommit job',
+  );
+
+  // Fetch the configuration necessary to submit a successful HTTPS request.
+  $sandcastle_config = getSandcastleConfig();
+
+  $app = $sandcastle_config[0];
+  $token = $sandcastle_config[1];
+
+  $cmd = 'curl -s -k -F app=' . $app . ' '
+          . '-F token=' . $token . ' -F job=\'' . json_encode($job)
+          .'\' "' . $url . '"';
+
+  $output = shell_exec($cmd);
+  assert(strlen($output) > 0);
+
+  // Extract Sandcastle URL from the response.
+  preg_match('/url": "(.+)"/', $output, $sandcastle_url);
+
+  assert(count($sandcastle_url) > 0, "Unable to submit Sandcastle request.");
+  assert(strlen($sandcastle_url[1]) > 0, "Unable to extract Sandcastle URL.");
+
+  if ($applyDiff) {
+    echo "\nSandcastle URL: " . $sandcastle_url[1] . "\n";
+    // Ask Phabricator to display it on the diff UI.
+    postURL($diffID, $sandcastle_url[1]);
+  } else {
+    echo "Continuous integration started Sandcastle tests. You can look at ";
+    echo "the progress at:\n" . $sandcastle_url[1] . "\n";
+  }
+}
+
+// Continuous run cript will set the environment variable and based on that
+// we'll trigger the execution of tests in Sandcastle. In that case we don't
+// need to apply any diffs and there's no associated workflow either.
+if (getenv(ENV_POST_RECEIVE_HOOK)) {
+  startTestsInSandcastle(
+    false /* $applyDiff */,
+    NULL /* $workflow */,
+    NULL /* $diffID */);
+}

arcanist_util/cpp_linter/ArcanistCpplintLinter.php

@@ -0,0 +1,88 @@
+<?php
+
+/**
+ * Uses google's cpplint.py to check code. RocksDB team forked this file from
+ * phabricator's /src/lint/linter/ArcanistCpplintLinter.php, and customized it
+ * for its own use.
+ *
+ * You can get it here:
+ * http://google-styleguide.googlecode.com/svn/trunk/cpplint/cpplint.py
+ * @group linter
+ */
+final class ArcanistCpplintLinter extends ArcanistLinter {
+
+  public function willLintPaths(array $paths) {
+    return;
+  }
+
+  public function getLinterName() {
+    return 'cpplint.py';
+  }
+
+  public function getLintPath() {
+    $bin = 'cpplint.py';
+    // Search under current dir
+    list($err) = exec_manual('which %s/%s', $this->linterDir(), $bin);
+    if (!$err) {
+      return $this->linterDir().'/'.$bin;
+    }
+
+    // Look for globally installed cpplint.py
+    list($err) = exec_manual('which %s', $bin);
+    if ($err) {
+      throw new ArcanistUsageException(
+        "cpplint.py does not appear to be installed on this system. Install ".
+        "it (e.g., with 'wget \"http://google-styleguide.googlecode.com/".
+        "svn/trunk/cpplint/cpplint.py\"') ".
+        "in your .arcconfig to point to the directory where it resides. ".
+        "Also don't forget to chmod a+x cpplint.py!");
+    }
+
+    return $bin;
+  }
+
+  public function lintPath($path) {
+    $bin = $this->getLintPath();
+    $path = $this->rocksdbDir().'/'.$path;
+
+    $f = new ExecFuture("%C $path", $bin);
+
+    list($err, $stdout, $stderr) = $f->resolve();
+
+    if ($err === 2) {
+      throw new Exception("cpplint failed to run correctly:\n".$stderr);
+    }
+
+    $lines = explode("\n", $stderr);
+    $messages = array();
+    foreach ($lines as $line) {
+      $line = trim($line);
+      $matches = null;
+      $regex = '/^[^:]+:(\d+):\s*(.*)\s*\[(.*)\] \[(\d+)\]$/';
+      if (!preg_match($regex, $line, $matches)) {
+        continue;
+      }
+      foreach ($matches as $key => $match) {
+        $matches[$key] = trim($match);
+      }
+      $message = new ArcanistLintMessage();
+      $message->setPath($path);
+      $message->setLine($matches[1]);
+      $message->setCode($matches[3]);
+      $message->setName($matches[3]);
+      $message->setDescription($matches[2]);
+      $message->setSeverity(ArcanistLintSeverity::SEVERITY_WARNING);
+      $this->addLintMessage($message);
+    }
+  }
+
+  // The path of this linter
+  private function linterDir() {
+    return dirname(__FILE__);
+  }
+
+  // TODO(kaili) a quick and dirty way to figure out rocksdb's root dir.
+  private function rocksdbDir() {
+    return $this->linterDir()."/../..";
+  }
+}

arcanist_util/cpp_linter/BaseDirectoryScopedFormatLinter.php

@@ -0,0 +1,74 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+abstract class BaseDirectoryScopedFormatLinter extends ArcanistLinter {
+
+  const LINT_FORMATTING = 1;
+
+  private $changedLines = array();
+  private $rawLintOutput = array();
+
+  abstract protected function getPathsToLint();
+
+  protected function shouldLintPath($path) {
+    foreach ($this->getPathsToLint() as $p) {
+      // check if $path starts with $p
+      if (strncmp($path, $p, strlen($p)) === 0) {
+        return true;
+      }
+    }
+    return false;
+  }
+
+  // API to tell this linter which lines were changed
+  final public function setPathChangedLines($path, $changed) {
+    $this->changedLines[$path] = $changed;
+  }
+
+  final public function willLintPaths(array $paths) {
+    $futures = array();
+    foreach ($paths as $path) {
+      if (!$this->shouldLintPath($path)) {
+        continue;
+      }
+
+      $changed = $this->changedLines[$path];
+      if (!isset($changed)) {
+        // do not run linter if there are no changes
+        continue;
+      }
+
+      $futures[$path] = $this->getFormatFuture($path, $changed);
+    }
+
+    foreach (id(new FutureIterator($futures))->limit(8) as $p => $f) {
+      $this->rawLintOutput[$p] = $f->resolvex();
+    }
+  }
+
+  abstract protected function getFormatFuture($path, array $changed);
+  abstract protected function getLintMessage($diff);
+
+  final public function lintPath($path) {
+    if (!isset($this->rawLintOutput[$path])) {
+      return;
+    }
+
+    list($new_content) = $this->rawLintOutput[$path];
+    $old_content = $this->getData($path);
+
+    if ($new_content != $old_content) {
+      $diff = ArcanistDiffUtils::renderDifferences($old_content, $new_content);
+      $this->raiseLintAtOffset(
+        0,
+        self::LINT_FORMATTING,
+        $this->getLintMessage($diff),
+        $old_content,
+        $new_content);
+    }
+  }
+
+}

arcanist_util/cpp_linter/FacebookHowtoevenLinter.php

@@ -0,0 +1,223 @@
+<?php
+// Copyright 2015-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+final class FacebookHowtoevenLinter extends ArcanistLinter {
+
+  const VERSION = 'fd9192f324c36d28136d14380f0b552a1385b59b';
+
+  private $parsedTargets = array();
+
+  public function getLinterName() {
+    return 'Howtoeven';
+  }
+
+  protected function getSeverity($code) {
+    $severities = array(
+      ArcanistLintSeverity::SEVERITY_DISABLED,
+      ArcanistLintSeverity::SEVERITY_ADVICE,
+      ArcanistLintSeverity::SEVERITY_WARNING,
+      ArcanistLintSeverity::SEVERITY_ERROR,
+    );
+    return idx($severities, $code, ArcanistLintSeverity::SEVERITY_WARNING);
+  }
+
+  public function willLintPaths(array $paths) {
+    // Cleanup previous runs.
+    $this->localExecx("rm -rf _build/_lint");
+
+    // Build compilation database.
+    $lintable_paths = $this->getLintablePaths($paths);
+    $interesting_paths = $this->getInterestingPaths($lintable_paths);
+
+    if (!$lintable_paths) {
+      return;
+    }
+
+    // Run lint.
+    try {
+      $this->localExecx(
+        "%C %C -p _build/dev/ %Ls",
+        $this->getBinaryPath(),
+        $this->getFilteredIssues(),
+        $lintable_paths);
+    } catch (CommandException $exception) {
+      PhutilConsole::getConsole()->writeErr($exception->getMessage());
+    }
+
+    // Load results.
+    $result = id(
+      new SQLite3(
+        $this->getProjectRoot().'/_build/_lint/lint.db',
+        SQLITE3_OPEN_READONLY))
+      ->query("SELECT * FROM raised_issues");
+
+    while ($issue = $result->fetchArray(SQLITE3_ASSOC)) {
+      // Skip issues not part of the linted file.
+      if (in_array($issue['file'], $interesting_paths)) {
+        $this->addLintMessage(id(new ArcanistLintMessage())
+          ->setPath($issue['file'])
+          ->setLine($issue['line'])
+          ->setChar($issue['column'])
+          ->setCode('Howtoeven')
+          ->setSeverity($this->getSeverity($issue['severity']))
+          ->setName('Hte-'.$issue['name'])
+          ->setDescription(
+            sprintf(
+              "%s\n\n%s",
+              ($issue['message']) ? $issue['message'] : $issue['description'],
+              $issue['explanation']))
+          ->setOriginalText(idx($issue, 'original', ''))
+          ->setReplacementText(idx($issue, 'replacement', '')));
+      }
+    }
+  }
+
+  public function lintPath($path) {
+  }
+
+  /**
+   * Get the paths that we know how to lint.
+   *
+   * The strategy is to first look whether there's an existing compilation
+   * database and use that if it's exhaustive. We generate our own only if
+   * necessary.
+   */
+  private function getLintablePaths($paths) {
+    // Replace headers with existing sources.
+    for ($i = 0; $i < count($paths); $i++) {
+      if (preg_match("/\.h$/", $paths[$i])) {
+        $header = preg_replace("/\.h$/", ".cpp", $paths[$i]);
+        if (file_exists($header)) {
+          $paths[$i] = $header;
+        }
+      }
+    }
+
+    // Check if database exists and is exhaustive.
+    $available_paths = $this->getAvailablePaths();
+    $lintable_paths = array_intersect($paths, $available_paths);
+    if ($paths === $lintable_paths) {
+      return $lintable_paths;
+    }
+
+    // Generate our own database.
+    $targets = $this->getTargetsFor($paths);
+    if (!$targets) {
+      PhutilConsole::getConsole()->writeErr(
+        "No build targets found for %s\n",
+        implode(', ', $paths));
+      return array();
+    }
+
+    $this->localExecx("./tools/build/bin/fbconfig.par -r %Ls", $targets);
+    $this->localExecx("./tools/build/bin/fbmake.par gen_cdb");
+
+    $available_paths = $this->getAvailablePaths();
+    $lintable_paths = array_intersect($paths, $available_paths);
+    if ($paths != $lintable_paths) {
+      PhutilConsole::getConsole()->writeErr(
+        "Can't lint %s\n",
+        implode(', ', array_diff($paths, $available_paths)));
+    }
+
+    // Return what we know how to lint.
+    return $lintable_paths;
+  }
+
+  /**
+   * Get the available paths in the current compilation database.
+   */
+  private function getAvailablePaths() {
+    $database_path = $this->getProjectRoot()
+      .'/_build/dev/compile_commands.json';
+    if (!file_exists($database_path)) {
+      return array();
+    }
+
+    $entries = json_decode(file_get_contents($database_path), true);
+    $paths = array();
+    foreach ($entries as $entry) {
+      $paths[] = $entry['file'];
+    }
+    return $paths;
+  }
+
+  /**
+   * Search for the targets directories for the given files.
+   */
+  private static function getTargetsFor($paths) {
+    $targets = array();
+    foreach ($paths as $path) {
+      while (($path = dirname($path)) !== '.') {
+        if (in_array('TARGETS', scandir($path))) {
+          $contents = file_get_contents($path.'/TARGETS');
+          if (strpos($contents, 'cpp_binary') !== false) {
+            $targets[] = $path;
+            break;
+          }
+        }
+      }
+    }
+    return array_unique($targets);
+  }
+
+  /**
+   * The paths that we actually want to report on.
+   */
+  private function getInterestingPaths($paths) {
+    $headers = array();
+    foreach ($paths as $path) {
+      $headers[] = preg_replace("/\.cpp$/", ".h", $path);
+    }
+    return array_merge($paths, $headers);
+  }
+
+  /**
+   * The path where the binary is located. Will return the current dewey binary
+   * unless the `HOWTOEVEN_BUILD` environment variable is set.
+   */
+  private function getBinaryPath() {
+    $path = sprintf(
+      "/mnt/dewey/fbcode/.commits/%s/builds/howtoeven/client",
+      self::VERSION);
+
+    $build = getenv('HOWTOEVEN_BUILD');
+    if ($build) {
+      $path = sprintf(
+        "./_build/%s/tools/howtoeven/client",
+        $build);
+      if (!file_exists($path)) {
+        PhutilConsole::getConsole()->writeErr(">> %s does not exist\n", $path);
+        exit(1);
+      }
+    }
+
+    return $path;
+  }
+
+  /**
+   * Execute the command in the root directory.
+   */
+  private function localExecx($command /* , ... */) {
+    $arguments = func_get_args();
+    return newv('ExecFuture', $arguments)
+      ->setCWD($this->getProjectRoot())
+      ->resolvex();
+  }
+
+  /**
+   * The root of the project.
+   */
+  private function getProjectRoot() {
+    return $this->getEngine()->getWorkingCopy()->getProjectRoot();
+  }
+
+  private function getFilteredIssues() {
+    $issues = getenv('HOWTOEVEN_ISSUES');
+    return ($issues) ? csprintf('-issues %s', $issues) : '';
+  }
+
+}

arcanist_util/cpp_linter/FbcodeClangFormatLinter.php

@@ -0,0 +1,58 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+final class FbcodeClangFormatLinter extends BaseDirectoryScopedFormatLinter {
+
+  const LINT_FORMATTING = 1;
+  const CLANG_FORMAT_BINARY = '/mnt/vol/engshare/admin/scripts/clang-format';
+
+  protected function getPathsToLint() {
+    return array('');
+  }
+
+  public function getLinterName() {
+    return 'CLANG_FORMAT';
+  }
+
+  public function getLintSeverityMap() {
+    return array(
+      self::LINT_FORMATTING => ArcanistLintSeverity::SEVERITY_ADVICE,
+    );
+  }
+
+  public function getLintNameMap() {
+    return array(
+      self::LINT_FORMATTING => pht('Changes are not clang-formatted'),
+    );
+  }
+
+  protected function getFormatFuture($path, array $changed) {
+    $args = "";
+    foreach ($changed as $key => $value) {
+      $args .= " --lines=$key:$key";
+    }
+
+    $binary = self::CLANG_FORMAT_BINARY;
+    if (!file_exists($binary)) {
+      // trust the $PATH
+      $binary = "clang-format";
+    }
+
+    return new ExecFuture(
+      "%s %s $args",
+      $binary,
+      $this->getEngine()->getFilePathOnDisk($path));
+  }
+
+  protected function getLintMessage($diff) {
+    $link_to_clang_format =
+      "[[ http://fburl.com/clang-format | clang-format ]]";
+    return <<<LINT_MSG
+Changes in this file were not formatted using $link_to_clang_format.
+Please run build_tools/format-diff.sh or `make format`
+LINT_MSG;
+  }
+}

arcanist_util/cpp_linter/FbcodeCppLinter.php

@@ -0,0 +1,126 @@
+<?php
+// Copyright 2004-present Facebook.  All rights reserved.
+
+class FbcodeCppLinter extends ArcanistLinter {
+  const FLINT      = "/home/engshare/tools/flint";
+  const LINT_ERROR   = 1;
+  const LINT_WARNING = 2;
+  const LINT_ADVICE  = 3;
+  const C_FLAG = "--c_mode=true";
+
+  private $rawLintOutput = array();
+
+  public function willLintPaths(array $paths) {
+    if (!file_exists(self::FLINT)) {
+      return;
+    }
+    $futures = array();
+    foreach ($paths as $p) {
+      $lpath = $this->getEngine()->getFilePathOnDisk($p);
+      $lpath_file = file($lpath);
+      if (preg_match('/\.(c)$/', $lpath) ||
+          preg_match('/-\*-.*Mode: C[; ].*-\*-/', $lpath_file[0]) ||
+          preg_match('/vim(:.*)*:\s*(set\s+)?filetype=c\s*:/', $lpath_file[0])
+          ) {
+        $futures[$p] = new ExecFuture("%s %s %s 2>&1",
+                           self::FLINT, self::C_FLAG,
+                           $this->getEngine()->getFilePathOnDisk($p));
+      } else {
+        $futures[$p] = new ExecFuture("%s %s 2>&1",
+          self::FLINT, $this->getEngine()->getFilePathOnDisk($p));
+      }
+    }
+
+    foreach (Futures($futures)->limit(8) as $p => $f) {
+      $this->rawLintOutput[$p] = $f->resolvex();
+    }
+
+    return;
+  }
+
+  public function getLinterName() {
+    return "FBCPP";
+  }
+
+  public function lintPath($path) {
+    $this->runCppLint($path);
+  }
+
+  private function runCppLint($path) {
+    $msgs = $this->getCppLintOutput($path);
+    foreach ($msgs as $m) {
+      $this->raiseLintAtLine($m['line'], 0, $m['severity'], $m['msg']);
+    }
+  }
+
+  private function adviseOnEachPattern(
+    $path,
+    $regex,
+    $message,
+    $lint_type = self::LINT_ADVICE,
+    $match_idx = 0) {
+      $file_data = $this->getData($path);
+      $matches = array();
+      if (!preg_match_all($regex, $file_data, $matches, PREG_OFFSET_CAPTURE)) {
+        return;
+      }
+
+      foreach ($matches[$match_idx] as $match) {
+        list($match_str, $offset) = $match;
+        $this->raiseLintAtOffset($offset, $lint_type, $message, $match_str);
+      }
+  }
+
+  public function getLintSeverityMap() {
+    return array(
+      self::LINT_WARNING => ArcanistLintSeverity::SEVERITY_WARNING,
+      self::LINT_ADVICE  => ArcanistLintSeverity::SEVERITY_ADVICE,
+      self::LINT_ERROR   => ArcanistLintSeverity::SEVERITY_ERROR
+    );
+  }
+
+  public function getLintNameMap() {
+    return array(
+      self::LINT_ADVICE   => "CppLint Advice",
+      self::LINT_WARNING  => "CppLint Warning",
+      self::LINT_ERROR    => "CppLint Error"
+    );
+  }
+
+  private function getCppLintOutput($path) {
+     if (!array_key_exists($path, $this->rawLintOutput)) {
+       return array();
+     }
+    list($output) = $this->rawLintOutput[$path];
+
+    $msgs = array();
+    $current = null;
+    $matches = array();
+    foreach (explode("\n", $output) as $line) {
+      if (preg_match('/.*?:(\d+):(.*)/', $line, $matches)) {
+        if ($current) {
+          $msgs[] = $current;
+        }
+        $line = $matches[1];
+        $text = $matches[2];
+        if (preg_match('/.*Warning.*/', $text)) {
+          $sev = self::LINT_WARNING;
+        } else if (preg_match('/.*Advice.*/', $text)) {
+          $sev = self::LINT_ADVICE;
+        } else {
+          $sev = self::LINT_ERROR;
+        }
+        $current = array('line'     => $line,
+                         'msg'      => $text,
+                         'severity' => $sev);
+      } else if ($current) {
+        $current['msg'] .= ' ' . $line;
+      }
+    }
+    if ($current) {
+      $msgs[] = $current;
+    }
+
+    return $msgs;
+  }
+}

arcanist_util/lint_engine/FacebookFbcodeLintEngine.php

@@ -0,0 +1,138 @@
+<?php
+// Copyright 2004-present Facebook.  All rights reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+class FacebookFbcodeLintEngine extends ArcanistLintEngine {
+
+  public function buildLinters() {
+    $linters = array();
+    $paths = $this->getPaths();
+
+    // Remove all deleted files, which are not checked by the
+    // following linters.
+    foreach ($paths as $key => $path) {
+      if (!Filesystem::pathExists($this->getFilePathOnDisk($path))) {
+        unset($paths[$key]);
+      }
+    }
+
+    $generated_linter = new ArcanistGeneratedLinter();
+    $linters[] = $generated_linter;
+
+    $nolint_linter = new ArcanistNoLintLinter();
+    $linters[] = $nolint_linter;
+
+    $text_linter = new ArcanistTextLinter();
+    $text_linter->setCustomSeverityMap(array(
+      ArcanistTextLinter::LINT_LINE_WRAP
+        => ArcanistLintSeverity::SEVERITY_ADVICE,
+    ));
+    $linters[] = $text_linter;
+
+    $java_text_linter = new ArcanistTextLinter();
+    $java_text_linter->setMaxLineLength(100);
+    $java_text_linter->setCustomSeverityMap(array(
+      ArcanistTextLinter::LINT_LINE_WRAP
+        => ArcanistLintSeverity::SEVERITY_ADVICE,
+    ));
+    $linters[] = $java_text_linter;
+
+    $python_linter = new ArcanistPEP8Linter();
+    $linters[] = $python_linter;
+
+    $cpp_linters = array();
+    $cpp_linters[] = $linters[] = new ArcanistCpplintLinter();
+    $cpp_linters[] = $linters[] = new FbcodeCppLinter();
+
+    $clang_format_linter = new FbcodeClangFormatLinter();
+    $linters[] = $clang_format_linter;
+
+    $spelling_linter = new ArcanistSpellingLinter();
+    $linters[] = $spelling_linter;
+
+    foreach ($paths as $path) {
+      $is_text = false;
+
+      $text_extensions = (
+        '/\.('.
+        'cpp|cxx|c|cc|h|hpp|hxx|tcc|'.
+        'py|rb|hs|pl|pm|tw|'.
+        'php|phpt|css|js|'.
+        'java|'.
+        'thrift|'.
+        'lua|'.
+        'siv|'.
+        'txt'.
+        ')$/'
+      );
+      if (preg_match($text_extensions, $path)) {
+        $is_text = true;
+      }
+      if ($is_text) {
+        $nolint_linter->addPath($path);
+
+        $generated_linter->addPath($path);
+        $generated_linter->addData($path, $this->loadData($path));
+
+        if (preg_match('/\.java$/', $path)) {
+          $java_text_linter->addPath($path);
+          $java_text_linter->addData($path, $this->loadData($path));
+        } else {
+          $text_linter->addPath($path);
+          $text_linter->addData($path, $this->loadData($path));
+        }
+
+        $spelling_linter->addPath($path);
+        $spelling_linter->addData($path, $this->loadData($path));
+      }
+      if (preg_match('/\.(cpp|c|cc|cxx|h|hh|hpp|hxx|tcc)$/', $path)
+          && !preg_match('/third-party/', $path)) {
+        foreach ($cpp_linters as &$linter) {
+          $linter->addPath($path);
+          $linter->addData($path, $this->loadData($path));
+        }
+
+        $clang_format_linter->addPath($path);
+        $clang_format_linter->addData($path, $this->loadData($path));
+        $clang_format_linter->setPathChangedLines(
+          $path, $this->getPathChangedLines($path));
+      }
+
+      // Match *.py and contbuild config files
+      if (preg_match('/(\.(py|tw|smcprops)|^contbuild\/configs\/[^\/]*)$/',
+                    $path)) {
+        $space_count = 4;
+        $real_path = $this->getFilePathOnDisk($path);
+        $dir = dirname($real_path);
+        do {
+          if (file_exists($dir.'/.python2space')) {
+            $space_count = 2;
+            break;
+          }
+          $dir = dirname($dir);
+        } while ($dir != '/' && $dir != '.');
+
+        $cur_path_linter = $python_linter;
+        $cur_path_linter->addPath($path);
+        $cur_path_linter->addData($path, $this->loadData($path));
+
+        if (preg_match('/\.tw$/', $path)) {
+          $cur_path_linter->setCustomSeverityMap(array(
+            'E251' => ArcanistLintSeverity::SEVERITY_DISABLED,
+          ));
+        }
+      }
+    }
+
+    $name_linter = new ArcanistFilenameLinter();
+    $linters[] = $name_linter;
+    foreach ($paths as $path) {
+      $name_linter->addPath($path);
+    }
+
+    return $linters;
+  }
+
+}

arcanist_util/lint_engine/FacebookHowtoevenLintEngine.php

@@ -0,0 +1,27 @@
+<?php
+// Copyright 2015-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+final class FacebookHowtoevenLintEngine extends ArcanistLintEngine {
+
+  public function buildLinters() {
+    $paths = array();
+
+    foreach ($this->getPaths() as $path) {
+      // Don't try to lint deleted files or changed directories.
+      if (!Filesystem::pathExists($path) || is_dir($path)) {
+        continue;
+      }
+
+      if (preg_match('/\.(cpp|c|cc|cxx|h|hh|hpp|hxx|tcc)$/', $path)) {
+        $paths[] = $path;
+      }
+    }
+
+    $howtoeven = new FacebookHowtoevenLinter();
+    $howtoeven->setPaths($paths);
+    return array($howtoeven);
+  }
+}

arcanist_util/unit_engine/FacebookFbcodeUnitTestEngine.php

@@ -0,0 +1,17 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+class FacebookFbcodeUnitTestEngine extends ArcanistUnitTestEngine {
+
+  public function run() {
+      // For a call to `arc call-conduit differential.updateunitresults` to
+      // succeed we need at least one entry here.
+      $result = new ArcanistUnitTestResult();
+      $result->setName("dummy_placeholder_entry");
+      $result->setResult(ArcanistUnitTestResult::RESULT_PASS);
+      return array($result);
+  }
+}

arcanist_util/unit_engine/FacebookOldFbcodeUnitTestEngine.php

@@ -0,0 +1,17 @@
+<?php
+// Copyright 2004-present Facebook. All Rights Reserved.
+// This source code is licensed under the BSD-style license found in the
+// LICENSE file in the root directory of this source tree. An additional grant
+// of patent rights can be found in the PATENTS file in the same directory.
+
+class FacebookFbcodeUnitTestEngine extends ArcanistBaseUnitTestEngine {
+
+  public function run() {
+      // For a call to `arc call-conduit differential.updateunitresults` to
+      // succeed we need at least one entry here.
+      $result = new ArcanistUnitTestResult();
+      $result->setName("dummy_placeholder_entry");
+      $result->setResult(ArcanistUnitTestResult::RESULT_PASS);
+      return array($result);
+  }
+}

build_tools/build_detect_platform

@@ -51,13 +51,11 @@ if [ -z "$ROCKSDB_NO_FBCODE" -a -d /mnt/gvfs/third-party ]; then
     FBCODE_BUILD="true"
     # If we're compiling with TSAN we need pic build
     PIC_BUILD=$COMPILE_WITH_TSAN
-    if [ -n "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
-      # we need this to build with MySQL. Don't use for other purposes.
-      source "$PWD/build_tools/fbcode_config4.8.1.sh"
-    elif [ -n "$ROCKSDB_FBCODE_BUILD_WITH_5xx" ]; then
+    if [ -z "$ROCKSDB_FBCODE_BUILD_WITH_481" ]; then
       source "$PWD/build_tools/fbcode_config.sh"
     else
-      source "$PWD/build_tools/fbcode_config_platform007.sh"
+      # we need this to build with MySQL. Don't use for other purposes.
+      source "$PWD/build_tools/fbcode_config4.8.1.sh"
     fi
 fi
 

build_tools/dependencies_platform007.sh

@@ -1,18 +0,0 @@
-GCC_BASE=/mnt/gvfs/third-party2/gcc/6e8e715624fd15256a7970073387793dfcf79b46/7.x/centos7-native/b2ef2b6
-CLANG_BASE=/mnt/gvfs/third-party2/llvm-fb/ef37e1faa1c29782abfac1ae65a291b9b7966f6d/stable/centos7-native/c9f9104
-LIBGCC_BASE=/mnt/gvfs/third-party2/libgcc/c67031f0f739ac61575a061518d6ef5038f99f90/7.x/platform007/5620abc
-GLIBC_BASE=/mnt/gvfs/third-party2/glibc/60d6f124a78798b73944f5ba87c2306ae3460153/2.26/platform007/f259413
-SNAPPY_BASE=/mnt/gvfs/third-party2/snappy/7f9bdaada18f59bc27ec2b0871eb8a6144343aef/1.1.3/platform007/ca4da3d
-ZLIB_BASE=/mnt/gvfs/third-party2/zlib/22c2d65676fb7c23cfa797c4f6937f38b026f3cf/1.2.8/platform007/ca4da3d
-BZIP2_BASE=/mnt/gvfs/third-party2/bzip2/dc49a21c5fceec6456a7a28a94dcd16690af1337/1.0.6/platform007/ca4da3d
-LZ4_BASE=/mnt/gvfs/third-party2/lz4/907b498203d297947f3bb70b9466f47e100f1873/r131/platform007/ca4da3d
-ZSTD_BASE=/mnt/gvfs/third-party2/zstd/3ee276cbacfad3074e3f07bf826ac47f06970f4e/1.3.5/platform007/15a3614
-GFLAGS_BASE=/mnt/gvfs/third-party2/gflags/0b9929d2588991c65a57168bf88aff2db87c5d48/2.2.0/platform007/ca4da3d
-JEMALLOC_BASE=/mnt/gvfs/third-party2/jemalloc/9c910d36d6235cc40e8ff559358f1833452300ca/master/platform007/5b0f53e
-NUMA_BASE=/mnt/gvfs/third-party2/numa/9cbf2460284c669ed19c3ccb200a71f7dd7e53c7/2.0.11/platform007/ca4da3d
-LIBUNWIND_BASE=/mnt/gvfs/third-party2/libunwind/bf3d7497fe4e6d007354f0adffa16ce3003f8338/1.3/platform007/6f3e0a9
-TBB_BASE=/mnt/gvfs/third-party2/tbb/ff4e0b093534704d8abab678a4fd7f5ea7b094c7/2018_U5/platform007/ca4da3d
-KERNEL_HEADERS_BASE=/mnt/gvfs/third-party2/kernel-headers/b5c4a61a5c483ba24722005ae07895971a2ac707/fb/platform007/da39a3e
-BINUTILS_BASE=/mnt/gvfs/third-party2/binutils/92ff90349e2f43ea0a8246d8b1cf17b6869013e3/2.29.1/centos7-native/da39a3e
-VALGRIND_BASE=/mnt/gvfs/third-party2/valgrind/f3f697a28122e6bcd513273dd9c1ff23852fc59f/3.13.0/platform007/ca4da3d
-LUA_BASE=/mnt/gvfs/third-party2/lua/f0cd714433206d5139df61659eb7b28b1dea6683/5.3.4/platform007/5007832

build_tools/fbcode_config_platform007.sh

@@ -1,157 +0,0 @@
-#!/bin/sh
-#
-# Set environment variables so that we can compile rocksdb using
-# fbcode settings.  It uses the latest g++ and clang compilers and also
-# uses jemalloc
-# Environment variables that change the behavior of this script:
-# PIC_BUILD -- if true, it will only take pic versions of libraries from fbcode. libraries that don't have pic variant will not be included
-
-
-BASEDIR=`dirname $BASH_SOURCE`
-source "$BASEDIR/dependencies_platform007.sh"
-
-CFLAGS=""
-
-# libgcc
-LIBGCC_INCLUDE="$LIBGCC_BASE/include/c++/7.3.0"
-LIBGCC_LIBS=" -L $LIBGCC_BASE/lib"
-
-# glibc
-GLIBC_INCLUDE="$GLIBC_BASE/include"
-GLIBC_LIBS=" -L $GLIBC_BASE/lib"
-
-# snappy
-SNAPPY_INCLUDE=" -I $SNAPPY_BASE/include/"
-if test -z $PIC_BUILD; then
-  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy.a"
-else
-  SNAPPY_LIBS=" $SNAPPY_BASE/lib/libsnappy_pic.a"
-fi
-CFLAGS+=" -DSNAPPY"
-
-if test -z $PIC_BUILD; then
-  # location of zlib headers and libraries
-  ZLIB_INCLUDE=" -I $ZLIB_BASE/include/"
-  ZLIB_LIBS=" $ZLIB_BASE/lib/libz.a"
-  CFLAGS+=" -DZLIB"
-
-  # location of bzip headers and libraries
-  BZIP_INCLUDE=" -I $BZIP2_BASE/include/"
-  BZIP_LIBS=" $BZIP2_BASE/lib/libbz2.a"
-  CFLAGS+=" -DBZIP2"
-
-  LZ4_INCLUDE=" -I $LZ4_BASE/include/"
-  LZ4_LIBS=" $LZ4_BASE/lib/liblz4.a"
-  CFLAGS+=" -DLZ4"
-fi
-
-ZSTD_INCLUDE=" -I $ZSTD_BASE/include/"
-if test -z $PIC_BUILD; then
-  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd.a"
-else
-  ZSTD_LIBS=" $ZSTD_BASE/lib/libzstd_pic.a"
-fi
-CFLAGS+=" -DZSTD"
-
-# location of gflags headers and libraries
-GFLAGS_INCLUDE=" -I $GFLAGS_BASE/include/"
-if test -z $PIC_BUILD; then
-  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags.a"
-else
-  GFLAGS_LIBS=" $GFLAGS_BASE/lib/libgflags_pic.a"
-fi
-CFLAGS+=" -DGFLAGS=gflags"
-
-# location of jemalloc
-JEMALLOC_INCLUDE=" -I $JEMALLOC_BASE/include/"
-JEMALLOC_LIB=" $JEMALLOC_BASE/lib/libjemalloc.a"
-
-if test -z $PIC_BUILD; then
-  # location of numa
-  NUMA_INCLUDE=" -I $NUMA_BASE/include/"
-  NUMA_LIB=" $NUMA_BASE/lib/libnuma.a"
-  CFLAGS+=" -DNUMA"
-
-  # location of libunwind
-  LIBUNWIND="$LIBUNWIND_BASE/lib/libunwind.a"
-fi
-
-# location of TBB
-TBB_INCLUDE=" -isystem $TBB_BASE/include/"
-if test -z $PIC_BUILD; then
-  TBB_LIBS="$TBB_BASE/lib/libtbb.a"
-else
-  TBB_LIBS="$TBB_BASE/lib/libtbb_pic.a"
-fi
-CFLAGS+=" -DTBB"
-
-# use Intel SSE support for checksum calculations
-export USE_SSE=1
-export PORTABLE=1
-
-BINUTILS="$BINUTILS_BASE/bin"
-AR="$BINUTILS/ar"
-
-DEPS_INCLUDE="$SNAPPY_INCLUDE $ZLIB_INCLUDE $BZIP_INCLUDE $LZ4_INCLUDE $ZSTD_INCLUDE $GFLAGS_INCLUDE $NUMA_INCLUDE $TBB_INCLUDE"
-
-STDLIBS="-L $GCC_BASE/lib64"
-
-CLANG_BIN="$CLANG_BASE/bin"
-CLANG_LIB="$CLANG_BASE/lib"
-CLANG_SRC="$CLANG_BASE/../../src"
-
-CLANG_ANALYZER="$CLANG_BIN/clang++"
-CLANG_SCAN_BUILD="$CLANG_SRC/llvm/tools/clang/tools/scan-build/bin/scan-build"
-
-if [ -z "$USE_CLANG" ]; then
-  # gcc
-  CC="$GCC_BASE/bin/gcc"
-  CXX="$GCC_BASE/bin/g++"
-
-  CFLAGS+=" -B$BINUTILS/gold"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  JEMALLOC=1
-else
-  # clang
-  CLANG_INCLUDE="$CLANG_LIB/clang/stable/include"
-  CC="$CLANG_BIN/clang"
-  CXX="$CLANG_BIN/clang++"
-
-  KERNEL_HEADERS_INCLUDE="$KERNEL_HEADERS_BASE/include"
-
-  CFLAGS+=" -B$BINUTILS/gold -nostdinc -nostdlib"
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x "
-  CFLAGS+=" -isystem $LIBGCC_BASE/include/c++/7.x/x86_64-facebook-linux "
-  CFLAGS+=" -isystem $GLIBC_INCLUDE"
-  CFLAGS+=" -isystem $LIBGCC_INCLUDE"
-  CFLAGS+=" -isystem $CLANG_INCLUDE"
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE/linux "
-  CFLAGS+=" -isystem $KERNEL_HEADERS_INCLUDE "
-  CFLAGS+=" -Wno-expansion-to-defined "
-  CXXFLAGS="-nostdinc++"
-fi
-
-CFLAGS+=" $DEPS_INCLUDE"
-CFLAGS+=" -DROCKSDB_PLATFORM_POSIX -DROCKSDB_LIB_IO_POSIX -DROCKSDB_FALLOCATE_PRESENT -DROCKSDB_MALLOC_USABLE_SIZE -DROCKSDB_RANGESYNC_PRESENT -DROCKSDB_SCHED_GETCPU_PRESENT -DROCKSDB_SUPPORT_THREAD_LOCAL -DHAVE_SSE42"
-CXXFLAGS+=" $CFLAGS"
-
-EXEC_LDFLAGS=" $SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $NUMA_LIB $TBB_LIBS"
-EXEC_LDFLAGS+=" -B$BINUTILS/gold"
-EXEC_LDFLAGS+=" -Wl,--dynamic-linker,/usr/local/fbcode/platform007/lib/ld.so"
-EXEC_LDFLAGS+=" $LIBUNWIND"
-EXEC_LDFLAGS+=" -Wl,-rpath=/usr/local/fbcode/platform007/lib"
-# required by libtbb
-EXEC_LDFLAGS+=" -ldl"
-
-PLATFORM_LDFLAGS="$LIBGCC_LIBS $GLIBC_LIBS $STDLIBS -lgcc -lstdc++"
-
-EXEC_LDFLAGS_SHARED="$SNAPPY_LIBS $ZLIB_LIBS $BZIP_LIBS $LZ4_LIBS $ZSTD_LIBS $GFLAGS_LIBS $TBB_LIBS"
-
-VALGRIND_VER="$VALGRIND_BASE/bin/"
-
-# lua not supported because it's on track for deprecation, I think
-LUA_PATH=
-LUA_LIB=
-
-export CC CXX AR CFLAGS CXXFLAGS EXEC_LDFLAGS EXEC_LDFLAGS_SHARED VALGRIND_VER JEMALLOC_LIB JEMALLOC_INCLUDE CLANG_ANALYZER CLANG_SCAN_BUILD LUA_PATH LUA_LIB

build_tools/make_new_version.sh

@@ -0,0 +1,53 @@
+#!/bin/bash
+#  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+#  This source code is licensed under the BSD-style license found in the
+#  LICENSE file in the root directory of this source tree. An additional grant
+#  of patent rights can be found in the PATENTS file in the same directory.
+
+set -e
+if [ -z "$GIT" ]
+then
+  GIT="git"
+fi
+
+# Print out the colored progress info so that it can be brainlessly
+# distinguished by users.
+function title() {
+  echo -e "\033[1;32m$*\033[0m"
+}
+
+usage="Create new RocksDB version and prepare it for the release process\n"
+usage+="USAGE: ./make_new_version.sh <version> [<remote>]\n"
+usage+="  version: specify a version without '.fb' suffix (e.g. 5.4).\n"
+usage+="  remote: name of the remote to push the branch to (default: origin)."
+
+# -- Pre-check
+if [[ $# < 1 ]]; then
+  echo -e $usage
+  exit 1
+fi
+
+ROCKSDB_VERSION=$1
+
+REMOTE="origin"
+if [[ $# > 1 ]]; then
+  REMOTE=$2
+fi
+
+GIT_BRANCH=`git rev-parse --abbrev-ref HEAD`
+echo $GIT_BRANCH
+
+if [ $GIT_BRANCH != "master" ]; then
+  echo "Error: Current branch is '$GIT_BRANCH', Please switch to master branch."
+  exit 1
+fi
+
+title "Adding new tag for this release ..."
+BRANCH="$ROCKSDB_VERSION.fb"
+$GIT checkout -b $BRANCH
+
+# Setting up the proxy for remote repo access
+title "Pushing new branch to remote repo ..."
+git push $REMOTE --set-upstream $BRANCH
+
+title "Branch $BRANCH is pushed to github;"

build_tools/rocksdb-lego-determinator

@@ -85,9 +85,8 @@ NON_SHM="TMPD=/tmp/rocksdb_test_tmp"
 GCC_481="ROCKSDB_FBCODE_BUILD_WITH_481=1"
 ASAN="COMPILE_WITH_ASAN=1"
 CLANG="USE_CLANG=1"
-# in gcc-5 there are known problems with TSAN like https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71090.
-# using platform007 gives us gcc-8 or higher which has that bug fixed.
-TSAN="ROCKSDB_FBCODE_BUILD_WITH_PLATFORM007=1 COMPILE_WITH_TSAN=1"
+LITE="OPT=\"-DROCKSDB_LITE -g\""
+TSAN="COMPILE_WITH_TSAN=1"
 UBSAN="COMPILE_WITH_UBSAN=1"
 DISABLE_JEMALLOC="DISABLE_JEMALLOC=1"
 HTTP_PROXY="https_proxy=http://fwdproxy.29.prn1:8080 http_proxy=http://fwdproxy.29.prn1:8080 ftp_proxy=http://fwdproxy.29.prn1:8080"
@@ -131,35 +130,6 @@ DISABLE_COMMANDS="[
     }
 ]"
 
-#
-# RocksDB unit test in parallel
-# Currently we always have noise in our parallel runs. This job is to help
-# manage the noise
-#
-PARALLEL_UNIT_TEST_COMMANDS="[
-    {
-        'name':'Rocksdb Parallel Unit Test',
-        'oncall':'$ONCALL',
-        'steps': [
-            $CLEANUP_ENV,
-            {
-                'name':'Build and test RocksDB debug version',
-                'shell':'$DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || $CONTRUN_NAME=punit_check $TASK_CREATION_TOOL || cat t/log-*',
-                'user':'root',
-                $PARSER
-            },
-            $CLEANUP_ENV,
-            {
-                'name':'Build and test RocksDB debug version under gcc-4.8.1',
-                'shell':'$GCC_481 $DEBUG make -j$(nproc) all && $SHM make check > /dev/null 2>&1 || $CONTRUN_NAME=punit_check_gcc481 $TASK_CREATION_TOOL || cat t/log-*',
-                'user':'root',
-                $PARSER
-            },
-        ],
-        $REPORT
-    }
-]"
-
 #
 # RocksDB unit test
 #
@@ -777,9 +747,6 @@ JAVA_BUILD_TEST_COMMANDS="[
 
 
 case $1 in
-  punit)
-    echo $PARALLEL_UNIT_TEST_COMMANDS
-    ;;
   unit)
     echo $UNIT_TEST_COMMANDS
     ;;

build_tools/update_dependencies.sh

@@ -52,45 +52,6 @@ function get_lib_base()
   log_variable $__res_var
 }
 
-###########################################################
-#                platform007 dependencies                 #
-###########################################################
-
-OUTPUT="$BASEDIR/dependencies_platform007.sh"
-
-rm -f "$OUTPUT"
-touch "$OUTPUT"
-
-echo "Writing dependencies to $OUTPUT"
-
-# Compilers locations
-GCC_BASE=`readlink -f $TP2_LATEST/gcc/7.x/centos7-native/*/`
-CLANG_BASE=`readlink -f $TP2_LATEST/llvm-fb/stable/centos7-native/*/`
-
-log_variable GCC_BASE
-log_variable CLANG_BASE
-
-# Libraries locations
-get_lib_base libgcc     7.x     platform007
-get_lib_base glibc      2.26    platform007
-get_lib_base snappy     LATEST  platform007
-get_lib_base zlib       LATEST  platform007
-get_lib_base bzip2      LATEST  platform007
-get_lib_base lz4        LATEST  platform007
-get_lib_base zstd       LATEST  platform007
-get_lib_base gflags     LATEST  platform007
-get_lib_base jemalloc   LATEST  platform007
-get_lib_base numa       LATEST  platform007
-get_lib_base libunwind  LATEST  platform007
-get_lib_base tbb        LATEST  platform007
-
-get_lib_base kernel-headers fb platform007
-get_lib_base binutils   LATEST centos7-native
-get_lib_base valgrind   LATEST platform007
-get_lib_base lua        5.3.4  platform007
-
-git diff $OUTPUT
-
 ###########################################################
 #                   5.x dependencies                      #
 ###########################################################

db/builder.cc

@@ -45,9 +45,9 @@ TableBuilder* NewTableBuilder(
         int_tbl_prop_collector_factories,
     uint32_t column_family_id, const std::string& column_family_name,
     WritableFileWriter* file, const CompressionType compression_type,
-    const CompressionOptions& compression_opts,
-    int level,
-    const std::string* compression_dict, const bool skip_filters) {
+    const CompressionOptions& compression_opts, int level,
+    const std::string* compression_dict, const bool skip_filters,
+    const uint64_t creation_time) {
   assert((column_family_id ==
           TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
          column_family_name.empty());
@@ -55,7 +55,7 @@ TableBuilder* NewTableBuilder(
       TableBuilderOptions(ioptions, internal_comparator,
                           int_tbl_prop_collector_factories, compression_type,
                           compression_opts, compression_dict, skip_filters,
-                          column_family_name, level),
+                          column_family_name, level, creation_time),
       column_family_id, file);
 }
 
@@ -74,7 +74,8 @@ Status BuildTable(
     const CompressionOptions& compression_opts, bool paranoid_file_checks,
     InternalStats* internal_stats, TableFileCreationReason reason,
     EventLogger* event_logger, int job_id, const Env::IOPriority io_priority,
-    TableProperties* table_properties, int level) {
+    TableProperties* table_properties, int level,
+    const uint64_t creation_time) {
   assert((column_family_id ==
           TablePropertiesCollectorFactory::Context::kUnknownColumnFamily) ==
          column_family_name.empty());
@@ -123,7 +124,8 @@ Status BuildTable(
       builder = NewTableBuilder(
           ioptions, internal_comparator, int_tbl_prop_collector_factories,
           column_family_id, column_family_name, file_writer.get(), compression,
-          compression_opts, level);
+          compression_opts, level, nullptr /* compression_dict */,
+          false /* skip_filters */, creation_time);
     }
 
     MergeHelper merge(env, internal_comparator.user_comparator(),

db/builder.h

@@ -48,10 +48,9 @@ TableBuilder* NewTableBuilder(
         int_tbl_prop_collector_factories,
     uint32_t column_family_id, const std::string& column_family_name,
     WritableFileWriter* file, const CompressionType compression_type,
-    const CompressionOptions& compression_opts,
-    int level,
+    const CompressionOptions& compression_opts, int level,
     const std::string* compression_dict = nullptr,
-    const bool skip_filters = false);
+    const bool skip_filters = false, const uint64_t creation_time = 0);
 
 // Build a Table file from the contents of *iter.  The generated file
 // will be named according to number specified in meta. On success, the rest of
@@ -77,6 +76,7 @@ extern Status BuildTable(
     InternalStats* internal_stats, TableFileCreationReason reason,
     EventLogger* event_logger = nullptr, int job_id = 0,
     const Env::IOPriority io_priority = Env::IO_HIGH,
-    TableProperties* table_properties = nullptr, int level = -1);
+    TableProperties* table_properties = nullptr, int level = -1,
+    const uint64_t creation_time = 0);
 
 }  // namespace rocksdb

db/c.cc

@@ -1372,24 +1372,23 @@ void rocksdb_writebatch_put_log_data(
   b->rep.PutLogData(Slice(blob, len));
 }
 
-class H : public WriteBatch::Handler {
- public:
-  void* state_;
-  void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
-  void (*deleted_)(void*, const char* k, size_t klen);
-  virtual void Put(const Slice& key, const Slice& value) override {
-    (*put_)(state_, key.data(), key.size(), value.data(), value.size());
-  }
-  virtual void Delete(const Slice& key) override {
-    (*deleted_)(state_, key.data(), key.size());
-  }
-};
-
 void rocksdb_writebatch_iterate(
     rocksdb_writebatch_t* b,
     void* state,
     void (*put)(void*, const char* k, size_t klen, const char* v, size_t vlen),
     void (*deleted)(void*, const char* k, size_t klen)) {
+  class H : public WriteBatch::Handler {
+   public:
+    void* state_;
+    void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
+    void (*deleted_)(void*, const char* k, size_t klen);
+    virtual void Put(const Slice& key, const Slice& value) override {
+      (*put_)(state_, key.data(), key.size(), value.data(), value.size());
+    }
+    virtual void Delete(const Slice& key) override {
+      (*deleted_)(state_, key.data(), key.size());
+    }
+  };
   H handler;
   handler.state_ = state;
   handler.put_ = put;
@@ -1634,6 +1633,18 @@ void rocksdb_writebatch_wi_iterate(
     void* state,
     void (*put)(void*, const char* k, size_t klen, const char* v, size_t vlen),
     void (*deleted)(void*, const char* k, size_t klen)) {
+  class H : public WriteBatch::Handler {
+   public:
+    void* state_;
+    void (*put_)(void*, const char* k, size_t klen, const char* v, size_t vlen);
+    void (*deleted_)(void*, const char* k, size_t klen);
+    virtual void Put(const Slice& key, const Slice& value) override {
+      (*put_)(state_, key.data(), key.size(), value.data(), value.size());
+    }
+    virtual void Delete(const Slice& key) override {
+      (*deleted_)(state_, key.data(), key.size());
+    }
+  };
   H handler;
   handler.state_ = state;
   handler.put_ = put;
@@ -2955,21 +2966,20 @@ void rocksdb_slicetransform_destroy(rocksdb_slicetransform_t* st) {
   delete st;
 }
 
-struct Wrapper : public rocksdb_slicetransform_t {
-  const SliceTransform* rep_;
-  ~Wrapper() { delete rep_; }
-  const char* Name() const override { return rep_->Name(); }
-  Slice Transform(const Slice& src) const override {
-    return rep_->Transform(src);
-  }
-  bool InDomain(const Slice& src) const override {
-    return rep_->InDomain(src);
-  }
-  bool InRange(const Slice& src) const override { return rep_->InRange(src); }
-  static void DoNothing(void*) { }
-};
-
 rocksdb_slicetransform_t* rocksdb_slicetransform_create_fixed_prefix(size_t prefixLen) {
+  struct Wrapper : public rocksdb_slicetransform_t {
+    const SliceTransform* rep_;
+    ~Wrapper() { delete rep_; }
+    const char* Name() const override { return rep_->Name(); }
+    Slice Transform(const Slice& src) const override {
+      return rep_->Transform(src);
+    }
+    bool InDomain(const Slice& src) const override {
+      return rep_->InDomain(src);
+    }
+    bool InRange(const Slice& src) const override { return rep_->InRange(src); }
+    static void DoNothing(void*) { }
+  };
   Wrapper* wrapper = new Wrapper;
   wrapper->rep_ = rocksdb::NewFixedPrefixTransform(prefixLen);
   wrapper->state_ = nullptr;
@@ -2978,6 +2988,19 @@ rocksdb_slicetransform_t* rocksdb_slicetransform_create_fixed_prefix(size_t pref
 }
 
 rocksdb_slicetransform_t* rocksdb_slicetransform_create_noop() {
+  struct Wrapper : public rocksdb_slicetransform_t {
+    const SliceTransform* rep_;
+    ~Wrapper() { delete rep_; }
+    const char* Name() const override { return rep_->Name(); }
+    Slice Transform(const Slice& src) const override {
+      return rep_->Transform(src);
+    }
+    bool InDomain(const Slice& src) const override {
+      return rep_->InDomain(src);
+    }
+    bool InRange(const Slice& src) const override { return rep_->InRange(src); }
+    static void DoNothing(void*) { }
+  };
   Wrapper* wrapper = new Wrapper;
   wrapper->rep_ = rocksdb::NewNoopTransform();
   wrapper->state_ = nullptr;

db/column_family.cc

@@ -353,8 +353,8 @@ ColumnFamilyData::ColumnFamilyData(
       initial_cf_options_(SanitizeOptions(db_options, cf_options)),
       ioptions_(db_options, initial_cf_options_),
       mutable_cf_options_(initial_cf_options_),
-      is_delete_range_supported_(strcmp(cf_options.table_factory->Name(),
-                                        BlockBasedTableFactory().Name()) == 0),
+      is_delete_range_supported_(
+          cf_options.table_factory->IsDeleteRangeSupported()),
       write_buffer_manager_(write_buffer_manager),
       mem_(nullptr),
       imm_(ioptions_.min_write_buffer_number_to_merge,

db/column_family_test.cc

@@ -302,6 +302,7 @@ class ColumnFamilyTest : public testing::Test {
         ASSERT_OK(Put(cf, key, RandomString(&rnd_, key_value_size - 10)));
       }
     }
+    db_->FlushWAL(false);
   }
 
 #ifndef ROCKSDB_LITE  // TEST functions in DB are not supported in lite
@@ -520,7 +521,6 @@ TEST_F(ColumnFamilyTest, DontReuseColumnFamilyID) {
   }
 }
 
-#ifndef ROCKSDB_LITE
 TEST_F(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
   Open();
 
@@ -542,7 +542,6 @@ TEST_F(ColumnFamilyTest, CreateCFRaceWithGetAggProperty) {
 
   rocksdb::SyncPoint::GetInstance()->DisableProcessing();
 }
-#endif  // !ROCKSDB_LITE
 
 class FlushEmptyCFTestWithParam : public ColumnFamilyTest,
                                   public testing::WithParamInterface<bool> {
@@ -581,6 +580,7 @@ TEST_P(FlushEmptyCFTestWithParam, FlushEmptyCFTest) {
   Flush(0);
   ASSERT_OK(Put(1, "bar", "v3"));  // seqID 4
   ASSERT_OK(Put(1, "foo", "v4"));  // seqID 5
+  db_->FlushWAL(false);
 
   // Preserve file system state up to here to simulate a crash condition.
   fault_env->SetFilesystemActive(false);
@@ -643,6 +643,7 @@ TEST_P(FlushEmptyCFTestWithParam, FlushEmptyCFTest2) {
   // Write to log file D
   ASSERT_OK(Put(1, "bar", "v4"));  // seqID 7
   ASSERT_OK(Put(1, "bar", "v5"));  // seqID 8
+  db_->FlushWAL(false);
   // Preserve file system state up to here to simulate a crash condition.
   fault_env->SetFilesystemActive(false);
   std::vector<std::string> names;

db/compaction.cc

@@ -283,32 +283,32 @@ bool Compaction::KeyNotExistsBeyondOutputLevel(
   assert(input_version_ != nullptr);
   assert(level_ptrs != nullptr);
   assert(level_ptrs->size() == static_cast<size_t>(number_levels_));
-  assert(cfd_->ioptions()->compaction_style != kCompactionStyleFIFO);
-  if (cfd_->ioptions()->compaction_style == kCompactionStyleUniversal) {
-    return bottommost_level_;
-  }
-  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel &&
-      output_level_ == 0) {
-    return false;
-  }
-  // Maybe use binary search to find right entry instead of linear search?
-  const Comparator* user_cmp = cfd_->user_comparator();
-  for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
-    const std::vector<FileMetaData*>& files = input_vstorage_->LevelFiles(lvl);
-    for (; level_ptrs->at(lvl) < files.size(); level_ptrs->at(lvl)++) {
-      auto* f = files[level_ptrs->at(lvl)];
-      if (user_cmp->Compare(user_key, f->largest.user_key()) <= 0) {
-        // We've advanced far enough
-        if (user_cmp->Compare(user_key, f->smallest.user_key()) >= 0) {
-          // Key falls in this file's range, so definitely
-          // exists beyond output level
-          return false;
+  if (cfd_->ioptions()->compaction_style == kCompactionStyleLevel) {
+    if (output_level_ == 0) {
+      return false;
+    }
+    // Maybe use binary search to find right entry instead of linear search?
+    const Comparator* user_cmp = cfd_->user_comparator();
+    for (int lvl = output_level_ + 1; lvl < number_levels_; lvl++) {
+      const std::vector<FileMetaData*>& files =
+          input_vstorage_->LevelFiles(lvl);
+      for (; level_ptrs->at(lvl) < files.size(); level_ptrs->at(lvl)++) {
+        auto* f = files[level_ptrs->at(lvl)];
+        if (user_cmp->Compare(user_key, f->largest.user_key()) <= 0) {
+          // We've advanced far enough
+          if (user_cmp->Compare(user_key, f->smallest.user_key()) >= 0) {
+            // Key falls in this file's range, so definitely
+            // exists beyond output level
+            return false;
+          }
+          break;
         }
-        break;
       }
     }
+    return true;
+  } else {
+    return bottommost_level_;
   }
-  return true;
 }
 
 // Mark (or clear) each file that is being compacted
@@ -463,4 +463,17 @@ bool Compaction::ShouldFormSubcompactions() const {
   }
 }
 
+uint64_t Compaction::MaxInputFileCreationTime() const {
+  uint64_t max_creation_time = 0;
+  for (const auto& file : inputs_[0].files) {
+    if (file->fd.table_reader != nullptr &&
+        file->fd.table_reader->GetTableProperties() != nullptr) {
+      uint64_t creation_time =
+          file->fd.table_reader->GetTableProperties()->creation_time;
+      max_creation_time = std::max(max_creation_time, creation_time);
+    }
+  }
+  return max_creation_time;
+}
+
 }  // namespace rocksdb

db/compaction.h

@@ -99,7 +99,8 @@ class Compaction {
   // input level.
   // REQUIREMENT: "compaction_input_level" must be >= 0 and
   //              < "input_levels()"
-  const std::vector<FileMetaData*>* inputs(size_t compaction_input_level) {
+  const std::vector<FileMetaData*>* inputs(
+      size_t compaction_input_level) const {
     assert(compaction_input_level < inputs_.size());
     return &inputs_[compaction_input_level].files;
   }
@@ -240,6 +241,8 @@ class Compaction {
 
   uint64_t max_compaction_bytes() const { return max_compaction_bytes_; }
 
+  uint64_t MaxInputFileCreationTime() const;
+
  private:
   // mark (or clear) all files that are being compacted
   void MarkFilesBeingCompacted(bool mark_as_compacted);

db/compaction_job.cc

@@ -1023,7 +1023,6 @@ Status CompactionJob::FinishCompactionOutputFile(
   uint64_t output_number = sub_compact->current_output()->meta.fd.GetNumber();
   assert(output_number != 0);
 
-  TableProperties table_properties;
   // Check for iterator errors
   Status s = input_status;
   auto meta = &sub_compact->current_output()->meta;
@@ -1079,6 +1078,23 @@ Status CompactionJob::FinishCompactionOutputFile(
   }
   sub_compact->outfile.reset();
 
+  if (s.ok() && current_entries == 0) {
+    // If there is nothing to output, no necessary to generate a sst file.
+    // This happens when the output level is bottom level, at the same time
+    // the sub_compact output nothing.
+    std::string fname = TableFileName(
+        db_options_.db_paths, meta->fd.GetNumber(), meta->fd.GetPathId());
+    env_->DeleteFile(fname);
+
+    // Also need to remove the file from outputs, or it will be added to the
+    // VersionEdit.
+    assert(!sub_compact->outputs.empty());
+    sub_compact->outputs.pop_back();
+    sub_compact->builder.reset();
+    sub_compact->current_output_file_size = 0;
+    return s;
+  }
+
   ColumnFamilyData* cfd = sub_compact->compaction->column_family_data();
   TableProperties tp;
   if (s.ok() && current_entries > 0) {
@@ -1131,12 +1147,22 @@ Status CompactionJob::FinishCompactionOutputFile(
                             meta->fd.GetPathId());
     sfm->OnAddFile(fn);
     if (sfm->IsMaxAllowedSpaceReached()) {
+      // TODO(ajkr): should we return OK() if max space was reached by the final
+      // compaction output file (similarly to how flush works when full)?
+      s = Status::IOError("Max allowed space was reached");
+      TEST_SYNC_POINT(
+          "CompactionJob::FinishCompactionOutputFile:"
+          "MaxAllowedSpaceReached");
       InstrumentedMutexLock l(db_mutex_);
       if (db_bg_error_->ok()) {
-        s = Status::IOError("Max allowed space was reached");
-        *db_bg_error_ = s;
-        TEST_SYNC_POINT(
-            "CompactionJob::FinishCompactionOutputFile:MaxAllowedSpaceReached");
+        Status new_bg_error = s;
+        // may temporarily unlock and lock the mutex.
+        EventHelpers::NotifyOnBackgroundError(
+            cfd->ioptions()->listeners, BackgroundErrorReason::kCompaction,
+            &new_bg_error, db_mutex_);
+        if (!new_bg_error.ok()) {
+          *db_bg_error_ = new_bg_error;
+        }
       }
     }
   }
@@ -1251,14 +1277,22 @@ Status CompactionJob::OpenCompactionOutputFile(
   // data is going to be found
   bool skip_filters =
       cfd->ioptions()->optimize_filters_for_hits && bottommost_level_;
+
+  uint64_t output_file_creation_time =
+      sub_compact->compaction->MaxInputFileCreationTime();
+  if (output_file_creation_time == 0) {
+    int64_t _current_time = 0;
+    db_options_.env->GetCurrentTime(&_current_time);  // ignore error
+    output_file_creation_time = static_cast<uint64_t>(_current_time);
+  }
+
   sub_compact->builder.reset(NewTableBuilder(
       *cfd->ioptions(), cfd->internal_comparator(),
       cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
       sub_compact->outfile.get(), sub_compact->compaction->output_compression(),
       cfd->ioptions()->compression_opts,
-      sub_compact->compaction->output_level(),
-      &sub_compact->compression_dict,
-      skip_filters));
+      sub_compact->compaction->output_level(), &sub_compact->compression_dict,
+      skip_filters, output_file_creation_time));
   LogFlush(db_options_.info_log);
   return s;
 }

db/compaction_job_test.cc

@@ -142,6 +142,7 @@ class CompactionJobTest : public testing::Test {
   }
 
   void SetLastSequence(const SequenceNumber sequence_number) {
+    versions_->SetLastToBeWrittenSequence(sequence_number + 1);
     versions_->SetLastSequence(sequence_number + 1);
   }
 
@@ -337,6 +338,24 @@ TEST_F(CompactionJobTest, SimpleDeletion) {
   RunCompaction({files}, expected_results);
 }
 
+TEST_F(CompactionJobTest, OutputNothing) {
+  NewDB();
+
+  auto file1 = mock::MakeMockFile({{KeyStr("a", 1U, kTypeValue), "val"}});
+
+  AddMockFile(file1);
+
+  auto file2 = mock::MakeMockFile({{KeyStr("a", 2U, kTypeDeletion), ""}});
+
+  AddMockFile(file2);
+
+  auto expected_results = mock::MakeMockFile();
+
+  SetLastSequence(4U);
+  auto files = cfd_->current()->storage_info()->LevelFiles(0);
+  RunCompaction({files}, expected_results);
+}
+
 TEST_F(CompactionJobTest, SimpleOverwrite) {
   NewDB();
 

db/compaction_picker.cc

@@ -1403,16 +1403,87 @@ bool FIFOCompactionPicker::NeedsCompaction(
   return vstorage->CompactionScore(kLevel0) >= 1;
 }
 
-Compaction* FIFOCompactionPicker::PickCompaction(
+namespace {
+uint64_t GetTotalFilesSize(
+    const std::vector<FileMetaData*>& files) {
+  uint64_t total_size = 0;
+  for (const auto& f : files) {
+    total_size += f->fd.file_size;
+  }
+  return total_size;
+}
+}  // anonymous namespace
+
+Compaction* FIFOCompactionPicker::PickTTLCompaction(
     const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
     VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
-  assert(vstorage->num_levels() == 1);
+  assert(ioptions_.compaction_options_fifo.ttl > 0);
+
   const int kLevel0 = 0;
   const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
-  uint64_t total_size = 0;
-  for (const auto& file : level_files) {
-    total_size += file->fd.file_size;
+  uint64_t total_size = GetTotalFilesSize(level_files);
+
+  int64_t _current_time;
+  auto status = ioptions_.env->GetCurrentTime(&_current_time);
+  if (!status.ok()) {
+    ROCKS_LOG_BUFFER(log_buffer,
+                     "[%s] FIFO compaction: Couldn't get current time: %s. "
+                     "Not doing compactions based on TTL. ",
+                     cf_name.c_str(), status.ToString().c_str());
+    return nullptr;
   }
+  const uint64_t current_time = static_cast<uint64_t>(_current_time);
+
+  std::vector<CompactionInputFiles> inputs;
+  inputs.emplace_back();
+  inputs[0].level = 0;
+
+  for (auto ritr = level_files.rbegin(); ritr != level_files.rend(); ++ritr) {
+    auto f = *ritr;
+    if (f->fd.table_reader != nullptr &&
+        f->fd.table_reader->GetTableProperties() != nullptr) {
+      auto creation_time =
+          f->fd.table_reader->GetTableProperties()->creation_time;
+      if (creation_time == 0 ||
+          creation_time >=
+              (current_time - ioptions_.compaction_options_fifo.ttl)) {
+        break;
+      }
+      total_size -= f->compensated_file_size;
+      inputs[0].files.push_back(f);
+    }
+  }
+
+  // Return a nullptr and proceed to size-based FIFO compaction if:
+  // 1. there are no files older than ttl OR
+  // 2. there are a few files older than ttl, but deleting them will not bring
+  //    the total size to be less than max_table_files_size threshold.
+  if (inputs[0].files.empty() ||
+      total_size > ioptions_.compaction_options_fifo.max_table_files_size) {
+    return nullptr;
+  }
+
+  for (const auto& f : inputs[0].files) {
+    ROCKS_LOG_BUFFER(log_buffer,
+                     "[%s] FIFO compaction: picking file %" PRIu64
+                     " with creation time %" PRIu64 " for deletion",
+                     cf_name.c_str(), f->fd.GetNumber(),
+                     f->fd.table_reader->GetTableProperties()->creation_time);
+  }
+
+  Compaction* c = new Compaction(
+      vstorage, ioptions_, mutable_cf_options, std::move(inputs), 0, 0, 0, 0,
+      kNoCompression, {}, /* is manual */ false, vstorage->CompactionScore(0),
+      /* is deletion compaction */ true, CompactionReason::kFIFOTtl);
+  return c;
+}
+
+Compaction* FIFOCompactionPicker::PickSizeCompaction(
+    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+    VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
+  const int kLevel0 = 0;
+  const std::vector<FileMetaData*>& level_files = vstorage->LevelFiles(kLevel0);
+  uint64_t total_size = GetTotalFilesSize(level_files);
 
   if (total_size <= ioptions_.compaction_options_fifo.max_table_files_size ||
       level_files.size() == 0) {
@@ -1433,7 +1504,6 @@ Compaction* FIFOCompactionPicker::PickCompaction(
             /* is manual */ false, vstorage->CompactionScore(0),
             /* is deletion compaction */ false,
             CompactionReason::kFIFOReduceNumFiles);
-        RegisterCompaction(c);
         return c;
       }
     }
@@ -1458,24 +1528,41 @@ Compaction* FIFOCompactionPicker::PickCompaction(
   std::vector<CompactionInputFiles> inputs;
   inputs.emplace_back();
   inputs[0].level = 0;
-  // delete old files (FIFO)
+
   for (auto ritr = level_files.rbegin(); ritr != level_files.rend(); ++ritr) {
     auto f = *ritr;
     total_size -= f->compensated_file_size;
     inputs[0].files.push_back(f);
     char tmp_fsize[16];
     AppendHumanBytes(f->fd.GetFileSize(), tmp_fsize, sizeof(tmp_fsize));
-    ROCKS_LOG_BUFFER(log_buffer, "[%s] FIFO compaction: picking file %" PRIu64
-                                 " with size %s for deletion",
+    ROCKS_LOG_BUFFER(log_buffer,
+                     "[%s] FIFO compaction: picking file %" PRIu64
+                     " with size %s for deletion",
                      cf_name.c_str(), f->fd.GetNumber(), tmp_fsize);
     if (total_size <= ioptions_.compaction_options_fifo.max_table_files_size) {
       break;
     }
   }
+
   Compaction* c = new Compaction(
       vstorage, ioptions_, mutable_cf_options, std::move(inputs), 0, 0, 0, 0,
       kNoCompression, {}, /* is manual */ false, vstorage->CompactionScore(0),
       /* is deletion compaction */ true, CompactionReason::kFIFOMaxSize);
+  return c;
+}
+
+Compaction* FIFOCompactionPicker::PickCompaction(
+    const std::string& cf_name, const MutableCFOptions& mutable_cf_options,
+    VersionStorageInfo* vstorage, LogBuffer* log_buffer) {
+  assert(vstorage->num_levels() == 1);
+
+  Compaction* c = nullptr;
+  if (ioptions_.compaction_options_fifo.ttl > 0) {
+    c = PickTTLCompaction(cf_name, mutable_cf_options, vstorage, log_buffer);
+  }
+  if (c == nullptr) {
+    c = PickSizeCompaction(cf_name, mutable_cf_options, vstorage, log_buffer);
+  }
   RegisterCompaction(c);
   return c;
 }

db/compaction_picker.h

@@ -242,6 +242,17 @@ class FIFOCompactionPicker : public CompactionPicker {
 
   virtual bool NeedsCompaction(
       const VersionStorageInfo* vstorage) const override;
+
+ private:
+  Compaction* PickTTLCompaction(const std::string& cf_name,
+                                const MutableCFOptions& mutable_cf_options,
+                                VersionStorageInfo* version,
+                                LogBuffer* log_buffer);
+
+  Compaction* PickSizeCompaction(const std::string& cf_name,
+                                 const MutableCFOptions& mutable_cf_options,
+                                 VersionStorageInfo* version,
+                                 LogBuffer* log_buffer);
 };
 
 class NullCompactionPicker : public CompactionPicker {

db/db_basic_test.cc

@@ -822,6 +822,9 @@ TEST_F(DBBasicTest, ChecksumTest) {
 // sense to run
 #ifndef OS_WIN
 TEST_F(DBBasicTest, MmapAndBufferOptions) {
+  if (!IsMemoryMappedAccessSupported()) {
+    return;
+  }
   Options options = CurrentOptions();
 
   options.use_direct_reads = true;

db/db_compaction_test.cc

@@ -1917,7 +1917,7 @@ TEST_F(DBCompactionTest, ManualAutoRace) {
 
   // The auto compaction is scheduled but waited until here
   TEST_SYNC_POINT("DBCompactionTest::ManualAutoRace:1");
-  // The auto compaction will wait until the the manual compaction is registerd
+  // The auto compaction will wait until the manual compaction is registerd
   // before processing so that it will be cancelled.
   dbfull()->CompactRange(CompactRangeOptions(), handles_[1], nullptr, nullptr);
   ASSERT_EQ("0,1", FilesPerLevel(1));

db/db_encryption_test.cc

@@ -0,0 +1,96 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//  This source code is also licensed under the GPLv2 license found in the
+//  COPYING file in the root directory of this source tree.
+//
+#include "db/db_test_util.h"
+#include "port/stack_trace.h"
+#include "rocksdb/perf_context.h"
+#if !defined(ROCKSDB_LITE)
+#include "util/sync_point.h"
+#endif
+#include <iostream>
+#include <string>
+
+namespace rocksdb {
+
+class DBEncryptionTest : public DBTestBase {
+ public:
+  DBEncryptionTest() : DBTestBase("/db_encryption_test") {}
+};
+
+#ifndef ROCKSDB_LITE
+
+TEST_F(DBEncryptionTest, CheckEncrypted) {
+  ASSERT_OK(Put("foo567", "v1.fetdq"));
+  ASSERT_OK(Put("bar123", "v2.dfgkjdfghsd"));
+  Close();
+
+  // Open all files and look for the values we've put in there.
+  // They should not be found if encrypted, otherwise
+  // they should be found.
+  std::vector<std::string> fileNames;
+  auto status = env_->GetChildren(dbname_, &fileNames);
+  ASSERT_OK(status);
+
+  auto defaultEnv = Env::Default();
+  int hits = 0;
+  for (auto it = fileNames.begin() ; it != fileNames.end(); ++it) {
+    if ((*it == "..") || (*it == ".")) {
+      continue;
+    }
+    auto filePath = dbname_ + "/" + *it;
+    unique_ptr<SequentialFile> seqFile;
+    auto envOptions = EnvOptions(CurrentOptions());
+    status = defaultEnv->NewSequentialFile(filePath, &seqFile, envOptions);
+    ASSERT_OK(status);
+
+    uint64_t fileSize;
+    status = defaultEnv->GetFileSize(filePath, &fileSize);
+    ASSERT_OK(status);
+
+    std::string scratch;
+    scratch.reserve(fileSize);
+    Slice data;
+    status = seqFile->Read(fileSize, &data, (char*)scratch.data());
+    ASSERT_OK(status);
+
+    if (data.ToString().find("foo567") != std::string::npos) {
+      hits++; 
+      //std::cout << "Hit in " << filePath << "\n";
+    }
+    if (data.ToString().find("v1.fetdq") != std::string::npos) {
+      hits++; 
+      //std::cout << "Hit in " << filePath << "\n";
+    }
+    if (data.ToString().find("bar123") != std::string::npos) {
+      hits++; 
+      //std::cout << "Hit in " << filePath << "\n";
+    }
+    if (data.ToString().find("v2.dfgkjdfghsd") != std::string::npos) {
+      hits++; 
+      //std::cout << "Hit in " << filePath << "\n";
+    }
+    if (data.ToString().find("dfgk") != std::string::npos) {
+      hits++; 
+      //std::cout << "Hit in " << filePath << "\n";
+    }
+  }
+  if (encrypted_env_) {
+    ASSERT_EQ(hits, 0);
+  } else {
+    ASSERT_GE(hits, 4);
+  }
+}
+
+#endif // ROCKSDB_LITE
+
+}  // namespace rocksdb
+
+int main(int argc, char** argv) {
+  rocksdb::port::InstallStackTraceHandler();
+  ::testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

db/db_impl.cc

@@ -158,6 +158,7 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       is_snapshot_supported_(true),
       write_buffer_manager_(immutable_db_options_.write_buffer_manager.get()),
       write_thread_(immutable_db_options_),
+      nonmem_write_thread_(immutable_db_options_),
       write_controller_(mutable_db_options_.delayed_write_rate),
       // Use delayed_write_rate as a base line to determine the initial
       // low pri write rate limit. It may be adjusted later.
@@ -187,7 +188,9 @@ DBImpl::DBImpl(const DBOptions& options, const std::string& dbname)
       bg_work_paused_(0),
       bg_compaction_paused_(0),
       refitting_level_(false),
-      opened_successfully_(false) {
+      opened_successfully_(false),
+      concurrent_prepare_(options.concurrent_prepare),
+      manual_wal_flush_(options.manual_wal_flush) {
   env_->GetAbsolutePath(dbname, &db_absolute_path_);
 
   // Reserve ten files or so for other uses and give the rest to TableCache.
@@ -610,6 +613,26 @@ int DBImpl::FindMinimumEmptyLevelFitting(ColumnFamilyData* cfd,
   return minimum_level;
 }
 
+Status DBImpl::FlushWAL(bool sync) {
+  {
+    // We need to lock log_write_mutex_ since logs_ might change concurrently
+    InstrumentedMutexLock wl(&log_write_mutex_);
+    log::Writer* cur_log_writer = logs_.back().writer;
+    auto s = cur_log_writer->WriteBuffer();
+    if (!s.ok()) {
+      ROCKS_LOG_ERROR(immutable_db_options_.info_log, "WAL flush error %s",
+                      s.ToString().c_str());
+    }
+    if (!sync) {
+      ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "FlushWAL sync=false");
+      return s;
+    }
+  }
+  // sync = true
+  ROCKS_LOG_DEBUG(immutable_db_options_.info_log, "FlushWAL sync=true");
+  return SyncWAL();
+}
+
 Status DBImpl::SyncWAL() {
   autovector<log::Writer*, 1> logs_to_sync;
   bool need_log_dir_sync;
@@ -648,6 +671,7 @@ Status DBImpl::SyncWAL() {
     need_log_dir_sync = !log_dir_synced_;
   }
 
+  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
   RecordTick(stats_, WAL_FILE_SYNCED);
   Status status;
   for (log::Writer* log : logs_to_sync) {
@@ -659,6 +683,7 @@ Status DBImpl::SyncWAL() {
   if (status.ok() && need_log_dir_sync) {
     status = directories_.GetWalDir()->Fsync();
   }
+  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
 
   TEST_SYNC_POINT("DBImpl::SyncWAL:BeforeMarkLogsSynced:1");
   {
@@ -2632,9 +2657,13 @@ Status DBImpl::IngestExternalFile(
     InstrumentedMutexLock l(&mutex_);
     TEST_SYNC_POINT("DBImpl::AddFile:MutexLock");
 
-    // Stop writes to the DB
+    // Stop writes to the DB by entering both write threads
     WriteThread::Writer w;
     write_thread_.EnterUnbatched(&w, &mutex_);
+    WriteThread::Writer nonmem_w;
+    if (concurrent_prepare_) {
+      nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+    }
 
     num_running_ingest_file_++;
 
@@ -2675,6 +2704,9 @@ Status DBImpl::IngestExternalFile(
     }
 
     // Resume writes to the DB
+    if (concurrent_prepare_) {
+      nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+    }
     write_thread_.ExitUnbatched(&w);
 
     // Update stats

db/db_impl.h

@@ -197,6 +197,7 @@ class DBImpl : public DB {
   using DB::Flush;
   virtual Status Flush(const FlushOptions& options,
                        ColumnFamilyHandle* column_family) override;
+  virtual Status FlushWAL(bool sync) override;
   virtual Status SyncWAL() override;
 
   virtual SequenceNumber GetLatestSequenceNumber() const override;
@@ -619,6 +620,10 @@ class DBImpl : public DB {
                             uint64_t* log_used = nullptr, uint64_t log_ref = 0,
                             bool disable_memtable = false);
 
+  Status WriteImplWALOnly(const WriteOptions& options, WriteBatch* updates,
+                          WriteCallback* callback = nullptr,
+                          uint64_t* log_used = nullptr, uint64_t log_ref = 0);
+
   uint64_t FindMinLogContainingOutstandingPrep();
   uint64_t FindMinPrepLogReferencedByMemTable();
 
@@ -744,12 +749,23 @@ class DBImpl : public DB {
   Status PreprocessWrite(const WriteOptions& write_options, bool* need_log_sync,
                          WriteContext* write_context);
 
+  WriteBatch* MergeBatch(const WriteThread::WriteGroup& write_group,
+                         WriteBatch* tmp_batch, size_t* write_with_wal);
+
+  Status WriteToWAL(const WriteBatch& merged_batch, log::Writer* log_writer,
+                    uint64_t* log_used, uint64_t* log_size);
+
   Status WriteToWAL(const WriteThread::WriteGroup& write_group,
-                    log::Writer* log_writer, bool need_log_sync,
-                    bool need_log_dir_sync, SequenceNumber sequence);
+                    log::Writer* log_writer, uint64_t* log_used,
+                    bool need_log_sync, bool need_log_dir_sync,
+                    SequenceNumber sequence);
+
+  Status ConcurrentWriteToWAL(const WriteThread::WriteGroup& write_group,
+                              uint64_t* log_used, SequenceNumber* last_sequence,
+                              int total_count);
 
   // Used by WriteImpl to update bg_error_ if paranoid check is enabled.
-  void ParanoidCheck(const Status& status);
+  void WriteCallbackStatusCheck(const Status& status);
 
   // Used by WriteImpl to update bg_error_ in case of memtable insert error.
   void MemTableInsertStatusCheck(const Status& memtable_insert_status);
@@ -825,11 +841,16 @@ class DBImpl : public DB {
   // Lock over the persistent DB state.  Non-nullptr iff successfully acquired.
   FileLock* db_lock_;
 
-  // The mutex for options file related operations.
-  // NOTE: should never acquire options_file_mutex_ and mutex_ at the
-  //       same time.
-  InstrumentedMutex options_files_mutex_;
+  // In addition to mutex_, log_write_mutex_ protected writes to logs_ and
+  // logfile_number_. With concurrent_prepare it also protects alive_log_files_,
+  // and log_empty_. Refer to the definition of each variable below for more
+  // details.
+  InstrumentedMutex log_write_mutex_;
   // State below is protected by mutex_
+  // With concurrent_prepare enabled, some of the variables that accessed during
+  // WriteToWAL need different synchronization: log_empty_, alive_log_files_,
+  // logs_, logfile_number_. Refer to the definition of each variable below for
+  // more description.
   mutable InstrumentedMutex mutex_;
 
   std::atomic<bool> shutting_down_;
@@ -843,10 +864,20 @@ class DBImpl : public DB {
   // * whenever there is an error in background purge, flush or compaction
   // * whenever num_running_ingest_file_ goes to 0.
   InstrumentedCondVar bg_cv_;
+  // Writes are protected by locking both mutex_ and log_write_mutex_, and reads
+  // must be under either mutex_ or log_write_mutex_. Since after ::Open,
+  // logfile_number_ is currently updated only in write_thread_, it can be read
+  // from the same write_thread_ without any locks.
   uint64_t logfile_number_;
   std::deque<uint64_t>
       log_recycle_files;  // a list of log files that we can recycle
   bool log_dir_synced_;
+  // Without concurrent_prepare, read and writes to log_empty_ are protected by
+  // mutex_. Since it is currently updated/read only in write_thread_, it can be
+  // accessed from the same write_thread_ without any locks. With
+  // concurrent_prepare writes, where it can be updated in different threads,
+  // read and writes are protected by log_write_mutex_ instead. This is to avoid
+  // expesnive mutex_ lock during WAL write, which update log_empty_.
   bool log_empty_;
   ColumnFamilyHandleImpl* default_cf_handle_;
   InternalStats* default_cf_internal_stats_;
@@ -881,14 +912,26 @@ class DBImpl : public DB {
     // true for some prefix of logs_
     bool getting_synced = false;
   };
+  // Without concurrent_prepare, read and writes to alive_log_files_ are
+  // protected by mutex_. However since back() is never popped, and push_back()
+  // is done only from write_thread_, the same thread can access the item
+  // reffered by back() without mutex_. With concurrent_prepare_, writes
+  // are protected by locking both mutex_ and log_write_mutex_, and reads must
+  // be under either mutex_ or log_write_mutex_.
   std::deque<LogFileNumberSize> alive_log_files_;
   // Log files that aren't fully synced, and the current log file.
   // Synchronization:
-  //  - push_back() is done from write thread with locked mutex_,
-  //  - pop_front() is done from any thread with locked mutex_,
+  //  - push_back() is done from write_thread_ with locked mutex_ and
+  //  log_write_mutex_
+  //  - pop_front() is done from any thread with locked mutex_ and
+  //  log_write_mutex_
+  //  - reads are done with either locked mutex_ or log_write_mutex_
   //  - back() and items with getting_synced=true are not popped,
-  //  - it follows that write thread with unlocked mutex_ can safely access
-  //    back() and items with getting_synced=true.
+  //  - The same thread that sets getting_synced=true will reset it.
+  //  - it follows that the object referred by back() can be safely read from
+  //  the write_thread_ without using mutex
+  //  - it follows that the items with getting_synced=true can be safely read
+  //  from the same thread that has set getting_synced=true
   std::deque<LogWriterNumber> logs_;
   // Signaled when getting_synced becomes false for some of the logs_.
   InstrumentedCondVar log_sync_cv_;
@@ -937,8 +980,10 @@ class DBImpl : public DB {
   WriteBufferManager* write_buffer_manager_;
 
   WriteThread write_thread_;
-
   WriteBatch tmp_batch_;
+  // The write thread when the writers have no memtable write. This will be used
+  // in 2PC to batch the prepares separately from the serial commit.
+  WriteThread nonmem_write_thread_;
 
   WriteController write_controller_;
 
@@ -946,6 +991,8 @@ class DBImpl : public DB {
 
   // Size of the last batch group. In slowdown mode, next write needs to
   // sleep if it uses up the quota.
+  // Note: This is to protect memtable and compaction. If the batch only writes
+  // to the WAL its size need not to be included in this.
   uint64_t last_batch_group_size_;
 
   FlushScheduler flush_scheduler_;
@@ -1188,6 +1235,11 @@ class DBImpl : public DB {
   bool MCOverlap(ManualCompaction* m, ManualCompaction* m1);
 
   size_t GetWalPreallocateBlockSize(uint64_t write_buffer_size) const;
+
+  // When set, we use a seprate queue for writes that dont write to memtable. In
+  // 2PC these are the writes at Prepare phase.
+  const bool concurrent_prepare_;
+  const bool manual_wal_flush_;
 };
 
 extern Options SanitizeOptions(const std::string& db,

db/db_impl_compaction_flush.cc

@@ -14,6 +14,7 @@
 #include <inttypes.h>
 
 #include "db/builder.h"
+#include "db/event_helpers.h"
 #include "monitoring/iostats_context_imp.h"
 #include "monitoring/perf_context_imp.h"
 #include "monitoring/thread_status_updater.h"
@@ -59,7 +60,14 @@ Status DBImpl::SyncClosedLogs(JobContext* job_context) {
     // "number < current_log_number".
     MarkLogsSynced(current_log_number - 1, true, s);
     if (!s.ok()) {
-      bg_error_ = s;
+      Status new_bg_error = s;
+      // may temporarily unlock and lock the mutex.
+      EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                            BackgroundErrorReason::kFlush,
+                                            &new_bg_error, &mutex_);
+      if (!new_bg_error.ok()) {
+        bg_error_ = new_bg_error;
+      }
       TEST_SYNC_POINT("DBImpl::SyncClosedLogs:Failed");
       return s;
     }
@@ -134,9 +142,16 @@ Status DBImpl::FlushMemTableToOutputFile(
 
   if (!s.ok() && !s.IsShutdownInProgress() &&
       immutable_db_options_.paranoid_checks && bg_error_.ok()) {
-    // if a bad error happened (not ShutdownInProgress) and paranoid_checks is
-    // true, mark DB read-only
-    bg_error_ = s;
+    Status new_bg_error = s;
+    // may temporarily unlock and lock the mutex.
+    EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                          BackgroundErrorReason::kFlush,
+                                          &new_bg_error, &mutex_);
+    if (!new_bg_error.ok()) {
+      // if a bad error happened (not ShutdownInProgress), paranoid_checks is
+      // true, and the error isn't handled by callback, mark DB read-only
+      bg_error_ = new_bg_error;
+    }
   }
   if (s.ok()) {
 #ifndef ROCKSDB_LITE
@@ -151,10 +166,17 @@ Status DBImpl::FlushMemTableToOutputFile(
           immutable_db_options_.db_paths[0].path, file_meta.fd.GetNumber());
       sfm->OnAddFile(file_path);
       if (sfm->IsMaxAllowedSpaceReached() && bg_error_.ok()) {
-        bg_error_ = Status::IOError("Max allowed space was reached");
+        Status new_bg_error = Status::IOError("Max allowed space was reached");
         TEST_SYNC_POINT_CALLBACK(
             "DBImpl::FlushMemTableToOutputFile:MaxAllowedSpaceReached",
-            &bg_error_);
+            &new_bg_error);
+        // may temporarily unlock and lock the mutex.
+        EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                              BackgroundErrorReason::kFlush,
+                                              &new_bg_error, &mutex_);
+        if (!new_bg_error.ok()) {
+          bg_error_ = new_bg_error;
+        }
       }
     }
 #endif  // ROCKSDB_LITE
@@ -565,7 +587,14 @@ Status DBImpl::CompactFilesImpl(
                    c->column_family_data()->GetName().c_str(),
                    job_context->job_id, status.ToString().c_str());
     if (immutable_db_options_.paranoid_checks && bg_error_.ok()) {
-      bg_error_ = status;
+      Status new_bg_error = status;
+      // may temporarily unlock and lock the mutex.
+      EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                            BackgroundErrorReason::kCompaction,
+                                            &new_bg_error, &mutex_);
+      if (!new_bg_error.ok()) {
+        bg_error_ = new_bg_error;
+      }
     }
   }
 
@@ -1197,7 +1226,6 @@ Status DBImpl::BackgroundFlush(bool* made_progress, JobContext* job_context,
 void DBImpl::BackgroundCallFlush() {
   bool made_progress = false;
   JobContext job_context(next_job_id_.fetch_add(1), true);
-  assert(bg_flush_scheduled_);
 
   TEST_SYNC_POINT("DBImpl::BackgroundCallFlush:start");
 
@@ -1205,6 +1233,7 @@ void DBImpl::BackgroundCallFlush() {
                        immutable_db_options_.info_log.get());
   {
     InstrumentedMutexLock l(&mutex_);
+    assert(bg_flush_scheduled_);
     num_running_flushes_++;
 
     auto pending_outputs_inserted_elem =
@@ -1623,7 +1652,14 @@ Status DBImpl::BackgroundCompaction(bool* made_progress,
     ROCKS_LOG_WARN(immutable_db_options_.info_log, "Compaction error: %s",
                    status.ToString().c_str());
     if (immutable_db_options_.paranoid_checks && bg_error_.ok()) {
-      bg_error_ = status;
+      Status new_bg_error = status;
+      // may temporarily unlock and lock the mutex.
+      EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                            BackgroundErrorReason::kCompaction,
+                                            &new_bg_error, &mutex_);
+      if (!new_bg_error.ok()) {
+        bg_error_ = new_bg_error;
+      }
     }
   }
 
@@ -1728,7 +1764,7 @@ bool DBImpl::HaveManualCompaction(ColumnFamilyData* cfd) {
     }
     if ((cfd == (*it)->cfd) && (!((*it)->in_progress || (*it)->done))) {
       // Allow automatic compaction if manual compaction is
-      // is in progress
+      // in progress
       return true;
     }
     it++;

db/db_impl_files.cc

@@ -250,7 +250,13 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
       }
       job_context->size_log_to_delete += earliest.size;
       total_log_size_ -= earliest.size;
+      if (concurrent_prepare_) {
+        log_write_mutex_.Lock();
+      }
       alive_log_files_.pop_front();
+      if (concurrent_prepare_) {
+        log_write_mutex_.Unlock();
+      }
       // Current log should always stay alive since it can't have
       // number < MinLogNumber().
       assert(alive_log_files_.size());
@@ -263,7 +269,10 @@ void DBImpl::FindObsoleteFiles(JobContext* job_context, bool force,
         continue;
       }
       logs_to_free_.push_back(log.ReleaseWriter());
-      logs_.pop_front();
+      {
+        InstrumentedMutexLock wl(&log_write_mutex_);
+        logs_.pop_front();
+      }
     }
     // Current log cannot be obsolete.
     assert(!logs_.empty());

db/db_impl_open.cc

@@ -16,6 +16,7 @@
 #include "db/builder.h"
 #include "options/options_helper.h"
 #include "rocksdb/wal_filter.h"
+#include "table/block_based_table_factory.h"
 #include "util/rate_limiter.h"
 #include "util/sst_file_manager_impl.h"
 #include "util/sync_point.h"
@@ -162,6 +163,19 @@ static Status ValidateOptions(
             "universal and level compaction styles. ");
       }
     }
+    if (cfd.options.compaction_options_fifo.ttl > 0) {
+      if (db_options.max_open_files != -1) {
+        return Status::NotSupported(
+            "FIFO Compaction with TTL is only supported when files are always "
+            "kept open (set max_open_files = -1). ");
+      }
+      if (cfd.options.table_factory->Name() !=
+          BlockBasedTableFactory().Name()) {
+        return Status::NotSupported(
+            "FIFO Compaction with TTL is only supported in "
+            "Block-Based Table format. ");
+      }
+    }
   }
 
   if (db_options.db_paths.size() > 4) {
@@ -723,6 +737,7 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
     auto last_sequence = *next_sequence - 1;
     if ((*next_sequence != kMaxSequenceNumber) &&
         (versions_->LastSequence() <= last_sequence)) {
+      versions_->SetLastToBeWrittenSequence(last_sequence);
       versions_->SetLastSequence(last_sequence);
     }
   }
@@ -794,9 +809,15 @@ Status DBImpl::RecoverLogFiles(const std::vector<uint64_t>& log_numbers,
   if (data_seen && !flushed) {
     // Mark these as alive so they'll be considered for deletion later by
     // FindObsoleteFiles()
+    if (concurrent_prepare_) {
+      log_write_mutex_.Lock();
+    }
     for (auto log_number : log_numbers) {
       alive_log_files_.push_back(LogFileNumberSize(log_number));
     }
+    if (concurrent_prepare_) {
+      log_write_mutex_.Unlock();
+    }
   }
 
   event_logger_.Log() << "job" << job_id << "event"
@@ -830,6 +851,11 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
         *cfd->GetLatestMutableCFOptions();
     bool paranoid_file_checks =
         cfd->GetLatestMutableCFOptions()->paranoid_file_checks;
+
+    int64_t _current_time = 0;
+    env_->GetCurrentTime(&_current_time);  // ignore error
+    const uint64_t current_time = static_cast<uint64_t>(_current_time);
+
     {
       mutex_.Unlock();
 
@@ -849,7 +875,8 @@ Status DBImpl::WriteLevel0TableForRecovery(int job_id, ColumnFamilyData* cfd,
           GetCompressionFlush(*cfd->ioptions(), mutable_cf_options),
           cfd->ioptions()->compression_opts, paranoid_file_checks,
           cfd->internal_stats(), TableFileCreationReason::kRecovery,
-          &event_logger_, job_id);
+          &event_logger_, job_id, Env::IO_HIGH, nullptr /* table_properties */,
+          -1 /* level */, current_time);
       LogFlush(immutable_db_options_.info_log);
       ROCKS_LOG_DEBUG(immutable_db_options_.info_log,
                       "[%s] [WriteLevel0TableForRecovery]"
@@ -960,14 +987,17 @@ Status DB::Open(const DBOptions& db_options, const std::string& dbname,
     if (s.ok()) {
       lfile->SetPreallocationBlockSize(
           impl->GetWalPreallocateBlockSize(max_write_buffer_size));
-      impl->logfile_number_ = new_log_number;
-      unique_ptr<WritableFileWriter> file_writer(
-          new WritableFileWriter(std::move(lfile), opt_env_options));
-      impl->logs_.emplace_back(
-          new_log_number,
-          new log::Writer(
-              std::move(file_writer), new_log_number,
-              impl->immutable_db_options_.recycle_log_file_num > 0));
+      {
+        InstrumentedMutexLock wl(&impl->log_write_mutex_);
+        impl->logfile_number_ = new_log_number;
+        unique_ptr<WritableFileWriter> file_writer(
+            new WritableFileWriter(std::move(lfile), opt_env_options));
+        impl->logs_.emplace_back(
+            new_log_number,
+            new log::Writer(
+                std::move(file_writer), new_log_number,
+                impl->immutable_db_options_.recycle_log_file_num > 0));
+      }
 
       // set column family handles
       for (auto cf : column_families) {
@@ -1001,8 +1031,14 @@ Status DB::Open(const DBOptions& db_options, const std::string& dbname,
         delete impl->InstallSuperVersionAndScheduleWork(
             cfd, nullptr, *cfd->GetLatestMutableCFOptions());
       }
+      if (impl->concurrent_prepare_) {
+        impl->log_write_mutex_.Lock();
+      }
       impl->alive_log_files_.push_back(
           DBImpl::LogFileNumberSize(impl->logfile_number_));
+      if (impl->concurrent_prepare_) {
+        impl->log_write_mutex_.Unlock();
+      }
       impl->DeleteObsoleteFiles();
       s = impl->directories_.GetDbDir()->Fsync();
     }

db/db_impl_readonly.cc

@@ -3,7 +3,6 @@
 //  COPYING file in the root directory) and Apache 2.0 License
 //  (found in the LICENSE.Apache file in the root directory).
 
-
 #include "db/db_impl_readonly.h"
 
 #include "db/compacted_db_impl.h"

db/db_impl_write.cc

@@ -12,6 +12,7 @@
 #define __STDC_FORMAT_MACROS
 #endif
 #include <inttypes.h>
+#include "db/event_helpers.h"
 #include "monitoring/perf_context_imp.h"
 #include "options/options_helper.h"
 #include "util/sync_point.h"
@@ -63,6 +64,10 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   if (my_batch == nullptr) {
     return Status::Corruption("Batch is nullptr!");
   }
+  if (concurrent_prepare_ && immutable_db_options_.enable_pipelined_write) {
+    return Status::NotSupported(
+        "pipelined_writes is not compatible with concurrent prepares");
+  }
 
   Status status;
   if (write_options.low_pri) {
@@ -72,6 +77,11 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     }
   }
 
+  if (concurrent_prepare_ && disable_memtable) {
+    return WriteImplWALOnly(write_options, my_batch, callback, log_used,
+                            log_ref);
+  }
+
   if (immutable_db_options_.enable_pipelined_write) {
     return PipelinedWriteImpl(write_options, my_batch, callback, log_used,
                               log_ref, disable_memtable);
@@ -130,14 +140,22 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   WriteContext write_context;
   WriteThread::WriteGroup write_group;
   bool in_parallel_group = false;
-  uint64_t last_sequence = versions_->LastSequence();
+  uint64_t last_sequence = kMaxSequenceNumber;
+  if (!concurrent_prepare_) {
+    last_sequence = versions_->LastSequence();
+  }
 
   mutex_.Lock();
 
   bool need_log_sync = !write_options.disableWAL && write_options.sync;
   bool need_log_dir_sync = need_log_sync && !log_dir_synced_;
-  status = PreprocessWrite(write_options, &need_log_sync, &write_context);
-  log::Writer* cur_log_writer = logs_.back().writer;
+  if (!concurrent_prepare_ || !disable_memtable) {
+    // With concurrent writes we do preprocess only in the write thread that
+    // also does write to memtable to avoid sync issue on shared data structure
+    // with the other thread
+    status = PreprocessWrite(write_options, &need_log_sync, &write_context);
+  }
+  log::Writer* log_writer = logs_.back().writer;
 
   mutex_.Unlock();
 
@@ -177,24 +195,24 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
       }
     }
 
-    const SequenceNumber current_sequence = last_sequence + 1;
-    last_sequence += total_count;
-
+    const bool concurrent_update = concurrent_prepare_;
     // Update stats while we are an exclusive group leader, so we know
     // that nobody else can be writing to these particular stats.
     // We're optimistic, updating the stats before we successfully
     // commit.  That lets us release our leader status early.
     auto stats = default_cf_internal_stats_;
-    stats->AddDBStats(InternalStats::NUMBER_KEYS_WRITTEN, total_count);
+    stats->AddDBStats(InternalStats::NUMBER_KEYS_WRITTEN, total_count,
+                      concurrent_update);
     RecordTick(stats_, NUMBER_KEYS_WRITTEN, total_count);
-    stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size);
+    stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size,
+                      concurrent_update);
     RecordTick(stats_, BYTES_WRITTEN, total_byte_size);
-    stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1);
+    stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1, concurrent_update);
     RecordTick(stats_, WRITE_DONE_BY_SELF);
     auto write_done_by_other = write_group.size - 1;
     if (write_done_by_other > 0) {
-      stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER,
-                        write_done_by_other);
+      stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER, write_done_by_other,
+                        concurrent_update);
       RecordTick(stats_, WRITE_DONE_BY_OTHER, write_done_by_other);
     }
     MeasureTime(stats_, BYTES_PER_WRITE, total_byte_size);
@@ -205,14 +223,27 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
 
     PERF_TIMER_STOP(write_pre_and_post_process_time);
 
-    if (status.ok() && !write_options.disableWAL) {
-      PERF_TIMER_GUARD(write_wal_time);
-      status = WriteToWAL(write_group, cur_log_writer, need_log_sync,
-                          need_log_dir_sync, current_sequence);
-      if (log_used != nullptr) {
-        *log_used = logfile_number_;
+    if (!concurrent_prepare_) {
+      if (status.ok() && !write_options.disableWAL) {
+        PERF_TIMER_GUARD(write_wal_time);
+        status = WriteToWAL(write_group, log_writer, log_used, need_log_sync,
+                            need_log_dir_sync, last_sequence + 1);
+      }
+    } else {
+      if (status.ok() && !write_options.disableWAL) {
+        PERF_TIMER_GUARD(write_wal_time);
+        // LastToBeWrittenSequence is increased inside WriteToWAL under
+        // wal_write_mutex_ to ensure ordered events in WAL
+        status = ConcurrentWriteToWAL(write_group, log_used, &last_sequence,
+                                      total_count);
+      } else {
+        // Otherwise we inc seq number for memtable writes
+        last_sequence = versions_->FetchAddLastToBeWrittenSequence(total_count);
       }
     }
+    assert(last_sequence != kMaxSequenceNumber);
+    const SequenceNumber current_sequence = last_sequence + 1;
+    last_sequence += total_count;
 
     if (status.ok()) {
       PERF_TIMER_GUARD(write_memtable_time);
@@ -253,13 +284,22 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   PERF_TIMER_START(write_pre_and_post_process_time);
 
   if (!w.CallbackFailed()) {
-    ParanoidCheck(status);
+    WriteCallbackStatusCheck(status);
   }
 
   if (need_log_sync) {
     mutex_.Lock();
     MarkLogsSynced(logfile_number_, need_log_dir_sync, status);
     mutex_.Unlock();
+    // Requesting sync with concurrent_prepare_ is expected to be very rare. We
+    // hance provide a simple implementation that is not necessarily efficient.
+    if (concurrent_prepare_) {
+      if (manual_wal_flush_) {
+        status = FlushWAL(true);
+      } else {
+        status = SyncWAL();
+      }
+    }
   }
 
   bool should_exit_batch_group = true;
@@ -269,7 +309,9 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
     should_exit_batch_group = write_thread_.CompleteParallelMemTableWriter(&w);
   }
   if (should_exit_batch_group) {
-    versions_->SetLastSequence(last_sequence);
+    if (status.ok()) {
+      versions_->SetLastSequence(last_sequence);
+    }
     MemTableInsertStatusCheck(w.status);
     write_thread_.ExitAsBatchGroupLeader(write_group, w.status);
   }
@@ -301,7 +343,7 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
     bool need_log_sync = !write_options.disableWAL && write_options.sync;
     bool need_log_dir_sync = need_log_sync && !log_dir_synced_;
     w.status = PreprocessWrite(write_options, &need_log_sync, &write_context);
-    log::Writer* cur_log_writer = logs_.back().writer;
+    log::Writer* log_writer = logs_.back().writer;
     mutex_.Unlock();
 
     // This can set non-OK status if callback fail.
@@ -349,12 +391,12 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
                           wal_write_group.size - 1);
         RecordTick(stats_, WRITE_DONE_BY_OTHER, wal_write_group.size - 1);
       }
-      w.status = WriteToWAL(wal_write_group, cur_log_writer, need_log_sync,
-                            need_log_dir_sync, current_sequence);
+      w.status = WriteToWAL(wal_write_group, log_writer, log_used,
+                            need_log_sync, need_log_dir_sync, current_sequence);
     }
 
     if (!w.CallbackFailed()) {
-      ParanoidCheck(w.status);
+      WriteCallbackStatusCheck(w.status);
     }
 
     if (need_log_sync) {
@@ -400,21 +442,109 @@ Status DBImpl::PipelinedWriteImpl(const WriteOptions& write_options,
   }
 
   assert(w.state == WriteThread::STATE_COMPLETED);
-  if (log_used != nullptr) {
-    *log_used = w.log_used;
-  }
-
   return w.FinalStatus();
 }
 
-void DBImpl::ParanoidCheck(const Status& status) {
+Status DBImpl::WriteImplWALOnly(const WriteOptions& write_options,
+                                WriteBatch* my_batch, WriteCallback* callback,
+                                uint64_t* log_used, uint64_t log_ref) {
+  Status status;
+  PERF_TIMER_GUARD(write_pre_and_post_process_time);
+  WriteThread::Writer w(write_options, my_batch, callback, log_ref,
+                        true /* disable_memtable */);
+  if (write_options.disableWAL) {
+    return status;
+  }
+  RecordTick(stats_, WRITE_WITH_WAL);
+  StopWatch write_sw(env_, immutable_db_options_.statistics.get(), DB_WRITE);
+
+  nonmem_write_thread_.JoinBatchGroup(&w);
+  assert(w.state != WriteThread::STATE_PARALLEL_MEMTABLE_WRITER);
+  if (w.state == WriteThread::STATE_COMPLETED) {
+    if (log_used != nullptr) {
+      *log_used = w.log_used;
+    }
+    return w.FinalStatus();
+  }
+  // else we are the leader of the write batch group
+  assert(w.state == WriteThread::STATE_GROUP_LEADER);
+  WriteContext write_context;
+  WriteThread::WriteGroup write_group;
+  uint64_t last_sequence;
+  nonmem_write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
+  // Note: no need to update last_batch_group_size_ here since the batch writes
+  // to WAL only
+
+  uint64_t total_byte_size = 0;
+  for (auto* writer : write_group) {
+    if (writer->CheckCallback(this)) {
+      total_byte_size = WriteBatchInternal::AppendedByteSize(
+          total_byte_size, WriteBatchInternal::ByteSize(writer->batch));
+    }
+  }
+
+  const bool concurrent_update = true;
+  // Update stats while we are an exclusive group leader, so we know
+  // that nobody else can be writing to these particular stats.
+  // We're optimistic, updating the stats before we successfully
+  // commit.  That lets us release our leader status early.
+  auto stats = default_cf_internal_stats_;
+  stats->AddDBStats(InternalStats::BYTES_WRITTEN, total_byte_size,
+                    concurrent_update);
+  RecordTick(stats_, BYTES_WRITTEN, total_byte_size);
+  stats->AddDBStats(InternalStats::WRITE_DONE_BY_SELF, 1, concurrent_update);
+  RecordTick(stats_, WRITE_DONE_BY_SELF);
+  auto write_done_by_other = write_group.size - 1;
+  if (write_done_by_other > 0) {
+    stats->AddDBStats(InternalStats::WRITE_DONE_BY_OTHER, write_done_by_other,
+                      concurrent_update);
+    RecordTick(stats_, WRITE_DONE_BY_OTHER, write_done_by_other);
+  }
+  MeasureTime(stats_, BYTES_PER_WRITE, total_byte_size);
+
+  PERF_TIMER_STOP(write_pre_and_post_process_time);
+
+  PERF_TIMER_GUARD(write_wal_time);
+  // LastToBeWrittenSequence is increased inside WriteToWAL under
+  // wal_write_mutex_ to ensure ordered events in WAL
+  status = ConcurrentWriteToWAL(write_group, log_used, &last_sequence,
+                                0 /*total_count*/);
+  if (status.ok() && write_options.sync) {
+    // Requesting sync with concurrent_prepare_ is expected to be very rare. We
+    // hance provide a simple implementation that is not necessarily efficient.
+    if (manual_wal_flush_) {
+      status = FlushWAL(true);
+    } else {
+      status = SyncWAL();
+    }
+  }
+  PERF_TIMER_START(write_pre_and_post_process_time);
+
+  if (!w.CallbackFailed()) {
+    WriteCallbackStatusCheck(status);
+  }
+  nonmem_write_thread_.ExitAsBatchGroupLeader(write_group, w.status);
+  if (status.ok()) {
+    status = w.FinalStatus();
+  }
+  return status;
+}
+
+void DBImpl::WriteCallbackStatusCheck(const Status& status) {
   // Is setting bg_error_ enough here?  This will at least stop
   // compaction and fail any further writes.
   if (immutable_db_options_.paranoid_checks && !status.ok() &&
       !status.IsBusy() && !status.IsIncomplete()) {
     mutex_.Lock();
     if (bg_error_.ok()) {
-      bg_error_ = status;  // stop compaction & fail any further writes
+      Status new_bg_error = status;
+      // may temporarily unlock and lock the mutex.
+      EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                            BackgroundErrorReason::kWriteCallback,
+                                            &new_bg_error, &mutex_);
+      if (!new_bg_error.ok()) {
+        bg_error_ = new_bg_error;  // stop compaction & fail any further writes
+      }
     }
     mutex_.Unlock();
   }
@@ -429,7 +559,14 @@ void DBImpl::MemTableInsertStatusCheck(const Status& status) {
   if (!status.ok()) {
     mutex_.Lock();
     assert(bg_error_.ok());
-    bg_error_ = status;
+    Status new_bg_error = status;
+    // may temporarily unlock and lock the mutex.
+    EventHelpers::NotifyOnBackgroundError(immutable_db_options_.listeners,
+                                          BackgroundErrorReason::kMemTable,
+                                          &new_bg_error, &mutex_);
+    if (!new_bg_error.ok()) {
+      bg_error_ = new_bg_error;  // stop compaction & fail any further writes
+    }
     mutex_.Unlock();
   }
 }
@@ -502,13 +639,12 @@ Status DBImpl::PreprocessWrite(const WriteOptions& write_options,
   return status;
 }
 
-Status DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
-                          log::Writer* log_writer, bool need_log_sync,
-                          bool need_log_dir_sync, SequenceNumber sequence) {
-  Status status;
-
+WriteBatch* DBImpl::MergeBatch(const WriteThread::WriteGroup& write_group,
+                               WriteBatch* tmp_batch, size_t* write_with_wal) {
+  assert(write_with_wal != nullptr);
+  assert(tmp_batch != nullptr);
   WriteBatch* merged_batch = nullptr;
-  size_t write_with_wal = 0;
+  *write_with_wal = 0;
   auto* leader = write_group.leader;
   if (write_group.size == 1 && leader->ShouldWriteToWAL() &&
       leader->batch->GetWalTerminationPoint().is_cleared()) {
@@ -516,31 +652,64 @@ Status DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
     // contains one batch, that batch should be written to the WAL,
     // and the batch is not wanting to be truncated
     merged_batch = leader->batch;
-    leader->log_used = logfile_number_;
-    write_with_wal = 1;
+    *write_with_wal = 1;
   } else {
     // WAL needs all of the batches flattened into a single batch.
     // We could avoid copying here with an iov-like AddRecord
     // interface
-    merged_batch = &tmp_batch_;
+    merged_batch = tmp_batch;
     for (auto writer : write_group) {
       if (writer->ShouldWriteToWAL()) {
         WriteBatchInternal::Append(merged_batch, writer->batch,
                                    /*WAL_only*/ true);
-        write_with_wal++;
+        (*write_with_wal)++;
       }
+    }
+  }
+  return merged_batch;
+}
+
+// When concurrent_prepare_ is disabled, this function is called from the only
+// write thread. Otherwise this must be called holding log_write_mutex_.
+Status DBImpl::WriteToWAL(const WriteBatch& merged_batch,
+                          log::Writer* log_writer, uint64_t* log_used,
+                          uint64_t* log_size) {
+  assert(log_size != nullptr);
+  Slice log_entry = WriteBatchInternal::Contents(&merged_batch);
+  *log_size = log_entry.size();
+  Status status = log_writer->AddRecord(log_entry);
+  if (log_used != nullptr) {
+    *log_used = logfile_number_;
+  }
+  total_log_size_ += log_entry.size();
+  // TODO(myabandeh): it might be unsafe to access alive_log_files_.back() here
+  // since alive_log_files_ might be modified concurrently
+  alive_log_files_.back().AddSize(log_entry.size());
+  log_empty_ = false;
+  return status;
+}
+
+Status DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
+                          log::Writer* log_writer, uint64_t* log_used,
+                          bool need_log_sync, bool need_log_dir_sync,
+                          SequenceNumber sequence) {
+  Status status;
+
+  size_t write_with_wal = 0;
+  WriteBatch* merged_batch =
+      MergeBatch(write_group, &tmp_batch_, &write_with_wal);
+  if (merged_batch == write_group.leader->batch) {
+    write_group.leader->log_used = logfile_number_;
+  } else if (write_with_wal > 1) {
+    for (auto writer : write_group) {
       writer->log_used = logfile_number_;
     }
   }
 
   WriteBatchInternal::SetSequence(merged_batch, sequence);
 
-  Slice log_entry = WriteBatchInternal::Contents(merged_batch);
-  status = log_writer->AddRecord(log_entry);
-  total_log_size_ += log_entry.size();
-  alive_log_files_.back().AddSize(log_entry.size());
-  log_empty_ = false;
-  uint64_t log_size = log_entry.size();
+  uint64_t log_size;
+  status = WriteToWAL(*merged_batch, log_writer, log_used, &log_size);
 
   if (status.ok() && need_log_sync) {
     StopWatch sw(env_, stats_, WAL_FILE_SYNC_MICROS);
@@ -582,6 +751,48 @@ Status DBImpl::WriteToWAL(const WriteThread::WriteGroup& write_group,
   return status;
 }
 
+Status DBImpl::ConcurrentWriteToWAL(const WriteThread::WriteGroup& write_group,
+                                    uint64_t* log_used,
+                                    SequenceNumber* last_sequence,
+                                    int total_count) {
+  Status status;
+
+  WriteBatch tmp_batch;
+  size_t write_with_wal = 0;
+  WriteBatch* merged_batch =
+      MergeBatch(write_group, &tmp_batch, &write_with_wal);
+
+  // We need to lock log_write_mutex_ since logs_ and alive_log_files might be
+  // pushed back concurrently
+  log_write_mutex_.Lock();
+  if (merged_batch == write_group.leader->batch) {
+    write_group.leader->log_used = logfile_number_;
+  } else if (write_with_wal > 1) {
+    for (auto writer : write_group) {
+      writer->log_used = logfile_number_;
+    }
+  }
+  *last_sequence = versions_->FetchAddLastToBeWrittenSequence(total_count);
+  auto sequence = *last_sequence + 1;
+  WriteBatchInternal::SetSequence(merged_batch, sequence);
+
+  log::Writer* log_writer = logs_.back().writer;
+  uint64_t log_size;
+  status = WriteToWAL(*merged_batch, log_writer, log_used, &log_size);
+  log_write_mutex_.Unlock();
+
+  if (status.ok()) {
+    const bool concurrent = true;
+    auto stats = default_cf_internal_stats_;
+    stats->AddDBStats(InternalStats::WAL_FILE_BYTES, log_size, concurrent);
+    RecordTick(stats_, WAL_FILE_BYTES, log_size);
+    stats->AddDBStats(InternalStats::WRITE_WITH_WAL, write_with_wal,
+                      concurrent);
+    RecordTick(stats_, WRITE_WITH_WAL, write_with_wal);
+  }
+  return status;
+}
+
 Status DBImpl::HandleWALFull(WriteContext* write_context) {
   mutex_.AssertHeld();
   assert(write_context != nullptr);
@@ -767,8 +978,9 @@ Status DBImpl::ThrottleLowPriWritesIfNeeded(const WriteOptions& write_options,
       // is that in case the write is heavy, low pri writes may never have
       // a chance to run. Now we guarantee we are still slowly making
       // progress.
-      write_controller_.low_pri_rate_limiter()->Request(my_batch->GetDataSize(),
-                                                        Env::IO_HIGH, nullptr);
+      write_controller_.low_pri_rate_limiter()->Request(
+          my_batch->GetDataSize(), Env::IO_HIGH, nullptr /* stats */,
+          RateLimiter::OpType::kWrite);
     }
   }
   return Status::OK();
@@ -808,6 +1020,13 @@ void DBImpl::NotifyOnMemTableSealed(ColumnFamilyData* cfd,
 // REQUIRES: this thread is currently at the front of the writer queue
 Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   mutex_.AssertHeld();
+  WriteThread::Writer nonmem_w;
+  if (concurrent_prepare_) {
+    // SwitchMemtable is a rare event. To simply the reasoning, we make sure
+    // that there is no concurrent thread writing to WAL.
+    nonmem_write_thread_.EnterUnbatched(&nonmem_w, &mutex_);
+  }
+
   unique_ptr<WritableFile> lfile;
   log::Writer* new_log = nullptr;
   MemTable* new_mem = nullptr;
@@ -821,7 +1040,13 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   // Attempt to switch to a new memtable and trigger flush of old.
   // Do this without holding the dbmutex lock.
   assert(versions_->prev_log_number() == 0);
+  if (concurrent_prepare_) {
+    log_write_mutex_.Lock();
+  }
   bool creating_new_log = !log_empty_;
+  if (concurrent_prepare_) {
+    log_write_mutex_.Unlock();
+  }
   uint64_t recycle_log_number = 0;
   if (creating_new_log && immutable_db_options_.recycle_log_file_num &&
       !log_recycle_files.empty()) {
@@ -877,9 +1102,9 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
         lfile->SetPreallocationBlockSize(preallocate_block_size);
         unique_ptr<WritableFileWriter> file_writer(
             new WritableFileWriter(std::move(lfile), opt_env_opt));
-        new_log =
-            new log::Writer(std::move(file_writer), new_log_number,
-                            immutable_db_options_.recycle_log_file_num > 0);
+        new_log = new log::Writer(
+            std::move(file_writer), new_log_number,
+            immutable_db_options_.recycle_log_file_num > 0, manual_wal_flush_);
       }
     }
 
@@ -906,15 +1131,25 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
     assert(creating_new_log);
     assert(!new_mem);
     assert(!new_log);
+    if (concurrent_prepare_) {
+      nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+    }
     return s;
   }
   if (creating_new_log) {
+    log_write_mutex_.Lock();
     logfile_number_ = new_log_number;
     assert(new_log != nullptr);
     log_empty_ = true;
     log_dir_synced_ = false;
+    if (!logs_.empty()) {
+      // Alway flush the buffer of the last log before switching to a new one
+      log::Writer* cur_log_writer = logs_.back().writer;
+      cur_log_writer->WriteBuffer();
+    }
     logs_.emplace_back(logfile_number_, new_log);
     alive_log_files_.push_back(LogFileNumberSize(logfile_number_));
+    log_write_mutex_.Unlock();
   }
   for (auto loop_cfd : *versions_->GetColumnFamilySet()) {
     // all this is just optimization to delete logs that
@@ -936,6 +1171,9 @@ Status DBImpl::SwitchMemtable(ColumnFamilyData* cfd, WriteContext* context) {
   cfd->SetMemtable(new_mem);
   context->superversions_to_free_.push_back(InstallSuperVersionAndScheduleWork(
       cfd, new_superversion, mutable_cf_options));
+  if (concurrent_prepare_) {
+    nonmem_write_thread_.ExitUnbatched(&nonmem_w);
+  }
   return s;
 }
 

db/db_io_failure_test.cc

@@ -250,7 +250,315 @@ TEST_F(DBIOFailureTest, PutFailsParanoid) {
   // the next put should NOT fail
   ASSERT_TRUE(s.ok());
 }
+#if !(defined NDEBUG) || !defined(OS_WIN)
+TEST_F(DBIOFailureTest, FlushSstRangeSyncError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.write_buffer_size = 256 * 1024 * 1024;
+  options.writable_file_max_buffer_size = 128 * 1024;
+  options.bytes_per_sync = 128 * 1024;
+  options.level0_file_num_compaction_trigger = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(10));
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
 
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+
+  std::atomic<int> range_sync_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::RangeSync", [&](void* arg) {
+        if (range_sync_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("range sync dummy error");
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  Random rnd(301);
+  std::string rnd_str =
+      RandomString(&rnd, static_cast<int>(options.bytes_per_sync / 2));
+  std::string rnd_str_512kb = RandomString(&rnd, 512 * 1024);
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  // First 1MB doesn't get range synced
+  ASSERT_OK(Put(1, "foo0_0", rnd_str_512kb));
+  ASSERT_OK(Put(1, "foo0_1", rnd_str_512kb));
+  ASSERT_OK(Put(1, "foo1_1", rnd_str));
+  ASSERT_OK(Put(1, "foo1_2", rnd_str));
+  ASSERT_OK(Put(1, "foo1_3", rnd_str));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  ASSERT_OK(Put(1, "foo3_1", rnd_str));
+  ASSERT_OK(Put(1, "foo3_2", rnd_str));
+  ASSERT_OK(Put(1, "foo3_3", rnd_str));
+  ASSERT_OK(Put(1, "foo4", "bar"));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+
+  // Following writes should fail as flush failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar", Get(1, "foo"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  ASSERT_GE(1, range_sync_called.load());
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar", Get(1, "foo"));
+}
+
+TEST_F(DBIOFailureTest, CompactSstRangeSyncError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.write_buffer_size = 256 * 1024 * 1024;
+  options.writable_file_max_buffer_size = 128 * 1024;
+  options.bytes_per_sync = 128 * 1024;
+  options.level0_file_num_compaction_trigger = 2;
+  options.target_file_size_base = 256 * 1024 * 1024;
+  options.disable_auto_compactions = true;
+  BlockBasedTableOptions table_options;
+  table_options.filter_policy.reset(NewBloomFilterPolicy(10));
+  options.table_factory.reset(NewBlockBasedTableFactory(table_options));
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+
+  Random rnd(301);
+  std::string rnd_str =
+      RandomString(&rnd, static_cast<int>(options.bytes_per_sync / 2));
+  std::string rnd_str_512kb = RandomString(&rnd, 512 * 1024);
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  // First 1MB doesn't get range synced
+  ASSERT_OK(Put(1, "foo0_0", rnd_str_512kb));
+  ASSERT_OK(Put(1, "foo0_1", rnd_str_512kb));
+  ASSERT_OK(Put(1, "foo1_1", rnd_str));
+  ASSERT_OK(Put(1, "foo1_2", rnd_str));
+  ASSERT_OK(Put(1, "foo1_3", rnd_str));
+  Flush(1);
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(1, "foo3_1", rnd_str));
+  ASSERT_OK(Put(1, "foo3_2", rnd_str));
+  ASSERT_OK(Put(1, "foo3_3", rnd_str));
+  ASSERT_OK(Put(1, "foo4", "bar"));
+  Flush(1);
+  dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+
+  std::atomic<int> range_sync_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::RangeSync", [&](void* arg) {
+        if (range_sync_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("range sync dummy error");
+        }
+      });
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(dbfull()->SetOptions(handles_[1],
+                                 {
+                                     {"disable_auto_compactions", "false"},
+                                 }));
+  dbfull()->TEST_WaitForCompact();
+
+  // Following writes should fail as flush failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar", Get(1, "foo"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+  ASSERT_GE(1, range_sync_called.load());
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar", Get(1, "foo"));
+}
+
+TEST_F(DBIOFailureTest, FlushSstCloseError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.level0_file_num_compaction_trigger = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(2));
+
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+  std::atomic<int> close_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::Close", [&](void* arg) {
+        if (close_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("close dummy error");
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(1, "foo1", "bar1"));
+  ASSERT_OK(Put(1, "foo", "bar2"));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+
+  // Following writes should fail as flush failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar2", Get(1, "foo"));
+  ASSERT_EQ("bar1", Get(1, "foo1"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar2", Get(1, "foo"));
+  ASSERT_EQ("bar1", Get(1, "foo1"));
+}
+
+TEST_F(DBIOFailureTest, CompactionSstCloseError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.level0_file_num_compaction_trigger = 2;
+  options.disable_auto_compactions = true;
+
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  ASSERT_OK(Put(1, "foo", "bar2"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  ASSERT_OK(Put(1, "foo", "bar3"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  dbfull()->TEST_WaitForCompact();
+
+  std::atomic<int> close_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::Close", [&](void* arg) {
+        if (close_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("close dummy error");
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  ASSERT_OK(dbfull()->SetOptions(handles_[1],
+                                 {
+                                     {"disable_auto_compactions", "false"},
+                                 }));
+  dbfull()->TEST_WaitForCompact();
+
+  // Following writes should fail as compaction failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar3", Get(1, "foo"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar3", Get(1, "foo"));
+}
+
+TEST_F(DBIOFailureTest, FlushSstSyncError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.use_fsync = false;
+  options.level0_file_num_compaction_trigger = 4;
+  options.memtable_factory.reset(new SpecialSkipListFactory(2));
+
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+  std::atomic<int> sync_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::Sync", [&](void* arg) {
+        if (sync_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("sync dummy error");
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(1, "foo1", "bar1"));
+  ASSERT_OK(Put(1, "foo", "bar2"));
+  dbfull()->TEST_WaitForFlushMemTable(handles_[1]);
+
+  // Following writes should fail as flush failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar2", Get(1, "foo"));
+  ASSERT_EQ("bar1", Get(1, "foo1"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar2", Get(1, "foo"));
+  ASSERT_EQ("bar1", Get(1, "foo1"));
+}
+
+TEST_F(DBIOFailureTest, CompactionSstSyncError) {
+  Options options = CurrentOptions();
+  options.env = env_;
+  options.create_if_missing = true;
+  options.error_if_exists = false;
+  options.paranoid_checks = true;
+  options.level0_file_num_compaction_trigger = 2;
+  options.disable_auto_compactions = true;
+  options.use_fsync = false;
+
+  DestroyAndReopen(options);
+  CreateAndReopenWithCF({"pikachu"}, options);
+  Status s;
+
+  ASSERT_OK(Put(1, "foo", "bar"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  ASSERT_OK(Put(1, "foo", "bar2"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  ASSERT_OK(Put(1, "foo", "bar3"));
+  ASSERT_OK(Put(1, "foo2", "bar"));
+  Flush(1);
+  dbfull()->TEST_WaitForCompact();
+
+  std::atomic<int> sync_called(0);
+  rocksdb::SyncPoint::GetInstance()->SetCallBack(
+      "SpecialEnv::SStableFile::Sync", [&](void* arg) {
+        if (sync_called.fetch_add(1) == 0) {
+          Status* st = static_cast<Status*>(arg);
+          *st = Status::IOError("close dummy error");
+        }
+      });
+
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+  ASSERT_OK(dbfull()->SetOptions(handles_[1],
+                                 {
+                                     {"disable_auto_compactions", "false"},
+                                 }));
+  dbfull()->TEST_WaitForCompact();
+
+  // Following writes should fail as compaction failed.
+  ASSERT_NOK(Put(1, "foo2", "bar3"));
+  ASSERT_EQ("bar3", Get(1, "foo"));
+
+  rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+  ReopenWithColumnFamilies({"default", "pikachu"}, options);
+  ASSERT_EQ("bar3", Get(1, "foo"));
+}
+#endif  // !(defined NDEBUG) || !defined(OS_WIN)
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_iter.cc

@@ -642,6 +642,7 @@ void DBIter::ReverseToBackward() {
     while (iter_->Valid() &&
            user_comparator_->Compare(ikey.user_key, saved_key_.GetUserKey()) >
                0) {
+      assert(ikey.sequence != kMaxSequenceNumber);
       if (ikey.sequence > sequence_) {
         PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
       } else {
@@ -954,6 +955,7 @@ void DBIter::FindPrevUserKey() {
         ++num_skipped;
       }
     }
+    assert(ikey.sequence != kMaxSequenceNumber);
     if (ikey.sequence > sequence_) {
       PERF_COUNTER_ADD(internal_recent_skipped_count, 1);
     } else {

db/db_log_iter_test.cc

@@ -116,6 +116,7 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorRace) {
       dbfull()->Flush(FlushOptions());
       Put("key4", DummyString(1024));
       ASSERT_EQ(dbfull()->GetLatestSequenceNumber(), 4U);
+      dbfull()->FlushWAL(false);
 
       {
         auto iter = OpenTransactionLogIter(0);
@@ -132,6 +133,7 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorRace) {
 
       // "key5" would be written in a new memtable and log
       Put("key5", DummyString(1024));
+      dbfull()->FlushWAL(false);
       {
         // this iter would miss "key4" if not fixed
         auto iter = OpenTransactionLogIter(0);
@@ -181,6 +183,7 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorCorruptedLog) {
       Put("key"+ToString(i), DummyString(10));
     }
     dbfull()->Flush(FlushOptions());
+    dbfull()->FlushWAL(false);
     // Corrupt this log to create a gap
     rocksdb::VectorLogPtr wal_files;
     ASSERT_OK(dbfull()->GetSortedWalFiles(wal_files));
@@ -194,6 +197,7 @@ TEST_F(DBTestXactLogIterator, TransactionLogIteratorCorruptedLog) {
 
     // Insert a new entry to a new log file
     Put("key1025", DummyString(10));
+    dbfull()->FlushWAL(false);
     // Try to read from the beginning. Should stop before the gap and read less
     // than 1025 entries
     auto iter = OpenTransactionLogIter(0);

db/db_options_test.cc

@@ -414,6 +414,7 @@ TEST_F(DBOptionsTest, DeleteObsoleteFilesPeriodChange) {
 TEST_F(DBOptionsTest, MaxOpenFilesChange) {
   SpecialEnv env(env_);
   Options options;
+  options.env = CurrentOptions().env;
   options.max_open_files = -1;
 
   Reopen(options);

db/db_range_del_test.cc

@@ -27,6 +27,9 @@ class DBRangeDelTest : public DBTestBase {
 // ROCKSDB_LITE
 #ifndef ROCKSDB_LITE
 TEST_F(DBRangeDelTest, NonBlockBasedTableNotSupported) {
+  if (!IsMemoryMappedAccessSupported()) {
+    return;
+  }
   Options opts = CurrentOptions();
   opts.table_factory.reset(new PlainTableFactory());
   opts.prefix_extractor.reset(NewNoopTransform());
@@ -816,8 +819,115 @@ TEST_F(DBRangeDelTest, TailingIteratorRangeTombstoneUnsupported) {
   }
   db_->ReleaseSnapshot(snapshot);
 }
+
 #endif  // !ROCKSDB_UBSAN_RUN
 
+TEST_F(DBRangeDelTest, SubcompactionHasEmptyDedicatedRangeDelFile) {
+  const int kNumFiles = 2, kNumKeysPerFile = 4;
+  Options options = CurrentOptions();
+  options.compression = kNoCompression;
+  options.disable_auto_compactions = true;
+  options.level0_file_num_compaction_trigger = kNumFiles;
+  options.max_subcompactions = 2;
+  options.num_levels = 2;
+  options.target_file_size_base = 4096;
+  Reopen(options);
+
+  // need a L1 file for subcompaction to be triggered
+  ASSERT_OK(
+      db_->Put(WriteOptions(), db_->DefaultColumnFamily(), Key(0), "val"));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  MoveFilesToLevel(1);
+
+  // put enough keys to fill up the first subcompaction, and later range-delete
+  // them so that the first subcompaction outputs no key-values. In that case
+  // it'll consider making an SST file dedicated to range deletions.
+  for (int i = 0; i < kNumKeysPerFile; ++i) {
+    ASSERT_OK(db_->Put(WriteOptions(), db_->DefaultColumnFamily(), Key(i),
+                       std::string(1024, 'a')));
+  }
+  ASSERT_OK(db_->Flush(FlushOptions()));
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(0),
+                             Key(kNumKeysPerFile)));
+
+  // the above range tombstone can be dropped, so that one alone won't cause a
+  // dedicated file to be opened. We can make one protected by snapshot that
+  // must be considered. Make its range outside the first subcompaction's range
+  // to exercise the tricky part of the code.
+  const Snapshot* snapshot = db_->GetSnapshot();
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
+                             Key(kNumKeysPerFile + 1),
+                             Key(kNumKeysPerFile + 2)));
+  ASSERT_OK(db_->Flush(FlushOptions()));
+
+  ASSERT_EQ(kNumFiles, NumTableFilesAtLevel(0));
+  ASSERT_EQ(1, NumTableFilesAtLevel(1));
+
+  db_->EnableAutoCompaction({db_->DefaultColumnFamily()});
+  dbfull()->TEST_WaitForCompact();
+  db_->ReleaseSnapshot(snapshot);
+}
+
+TEST_F(DBRangeDelTest, MemtableBloomFilter) {
+  // regression test for #2743. the range delete tombstones in memtable should
+  // be added even when Get() skips searching due to its prefix bloom filter
+  const int kMemtableSize = 1 << 20;              // 1MB
+  const int kMemtablePrefixFilterSize = 1 << 13;  // 8KB
+  const int kNumKeys = 1000;
+  const int kPrefixLen = 8;
+  Options options = CurrentOptions();
+  options.memtable_prefix_bloom_size_ratio =
+      static_cast<double>(kMemtablePrefixFilterSize) / kMemtableSize;
+  options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(kPrefixLen));
+  options.write_buffer_size = kMemtableSize;
+  Reopen(options);
+
+  for (int i = 0; i < kNumKeys; ++i) {
+    ASSERT_OK(Put(Key(i), "val"));
+  }
+  Flush();
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(), Key(0),
+                             Key(kNumKeys)));
+  for (int i = 0; i < kNumKeys; ++i) {
+    std::string value;
+    ASSERT_TRUE(db_->Get(ReadOptions(), Key(i), &value).IsNotFound());
+  }
+}
+
+TEST_F(DBRangeDelTest, UnorderedTombstones) {
+  // Regression test for #2752. Range delete tombstones between
+  // different snapshot stripes are not stored in order, so the first
+  // tombstone of each snapshot stripe should be checked as a smallest
+  // candidate.
+  Options options = CurrentOptions();
+  DestroyAndReopen(options);
+
+  auto cf = db_->DefaultColumnFamily();
+
+  ASSERT_OK(db_->Put(WriteOptions(), cf, "a", "a"));
+  ASSERT_OK(db_->Flush(FlushOptions(), cf));
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+  ASSERT_OK(dbfull()->TEST_CompactRange(0, nullptr, nullptr));
+  ASSERT_EQ(1, NumTableFilesAtLevel(1));
+
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), cf, "b", "c"));
+  // Hold a snapshot to separate these two delete ranges.
+  auto snapshot = db_->GetSnapshot();
+  ASSERT_OK(db_->DeleteRange(WriteOptions(), cf, "a", "b"));
+  ASSERT_OK(db_->Flush(FlushOptions(), cf));
+  db_->ReleaseSnapshot(snapshot);
+
+  std::vector<std::vector<FileMetaData>> files;
+  dbfull()->TEST_GetFilesMetaData(cf, &files);
+  ASSERT_EQ(1, files[0].size());
+  ASSERT_EQ("a", files[0][0].smallest.user_key());
+  ASSERT_EQ("c", files[0][0].largest.user_key());
+
+  std::string v;
+  auto s = db_->Get(ReadOptions(), "a", &v);
+  ASSERT_TRUE(s.IsNotFound());
+}
+
 #endif  // ROCKSDB_LITE
 
 }  // namespace rocksdb

db/db_sst_test.cc

@@ -335,6 +335,7 @@ TEST_F(DBSSTTest, RateLimitedDelete) {
   ASSERT_OK(s);
   options.sst_file_manager->SetDeleteRateBytesPerSecond(rate_bytes_per_sec);
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
 
   ASSERT_OK(TryReopen(options));
   // Create 4 files in L0
@@ -400,6 +401,7 @@ TEST_F(DBSSTTest, DeleteSchedulerMultipleDBPaths) {
       env_, nullptr, trash_dir, rate_bytes_per_sec, false, &s));
   ASSERT_OK(s);
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
 
   DestroyAndReopen(options);
 
@@ -454,9 +456,14 @@ TEST_F(DBSSTTest, DestroyDBWithRateLimitedDelete) {
       [&](void* arg) { bg_delete_file++; });
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
 
+  Status s;
   Options options = CurrentOptions();
   options.disable_auto_compactions = true;
   options.env = env_;
+  std::string trash_dir = test::TmpDir(env_) + "/trash";
+  options.sst_file_manager.reset(
+      NewSstFileManager(env_, nullptr, trash_dir, 0, false, &s));
+  ASSERT_OK(s);
   DestroyAndReopen(options);
 
   // Create 4 files in L0
@@ -469,15 +476,12 @@ TEST_F(DBSSTTest, DestroyDBWithRateLimitedDelete) {
 
   // Close DB and destroy it using DeleteScheduler
   Close();
-  std::string trash_dir = test::TmpDir(env_) + "/trash";
-  int64_t rate_bytes_per_sec = 1024 * 1024;  // 1 Mb / Sec
-  Status s;
-  options.sst_file_manager.reset(NewSstFileManager(
-      env_, nullptr, trash_dir, rate_bytes_per_sec, false, &s));
-  ASSERT_OK(s);
-  ASSERT_OK(DestroyDB(dbname_, options));
 
   auto sfm = static_cast<SstFileManagerImpl*>(options.sst_file_manager.get());
+
+  sfm->SetDeleteRateBytesPerSecond(1024 * 1024);
+  sfm->delete_scheduler()->TEST_SetMaxTrashDBRatio(1.1);
+  ASSERT_OK(DestroyDB(dbname_, options));
   sfm->WaitForEmptyTrash();
   // We have deleted the 4 sst files in the delete_scheduler
   ASSERT_EQ(bg_delete_file, 4);

db/db_test.cc

@@ -2790,7 +2790,7 @@ TEST_F(DBTest, FIFOCompactionTestWithCompaction) {
   ASSERT_EQ(NumTableFilesAtLevel(0), 10);
 
   for (int i = 0; i < 60; i++) {
-    // Generate and flush a file about 10KB.
+    // Generate and flush a file about 20KB.
     for (int j = 0; j < 20; j++) {
       ASSERT_OK(Put(ToString(i * 20 + j + 2000), RandomString(&rnd, 980)));
     }
@@ -2805,6 +2805,285 @@ TEST_F(DBTest, FIFOCompactionTestWithCompaction) {
   ASSERT_LE(SizeAtLevel(0),
             options.compaction_options_fifo.max_table_files_size);
 }
+
+TEST_F(DBTest, FIFOCompactionStyleWithCompactionAndDelete) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.write_buffer_size = 20 << 10;  // 20K
+  options.arena_block_size = 4096;
+  options.compaction_options_fifo.max_table_files_size = 1500 << 10;  // 1MB
+  options.compaction_options_fifo.allow_compaction = true;
+  options.level0_file_num_compaction_trigger = 3;
+  options.compression = kNoCompression;
+  options.create_if_missing = true;
+  options = CurrentOptions(options);
+  DestroyAndReopen(options);
+
+  Random rnd(301);
+  for (int i = 0; i < 3; i++) {
+    // Each file contains a different key which will be dropped later.
+    ASSERT_OK(Put("a" + ToString(i), RandomString(&rnd, 500)));
+    ASSERT_OK(Put("key" + ToString(i), ""));
+    ASSERT_OK(Put("z" + ToString(i), RandomString(&rnd, 500)));
+    Flush();
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  }
+  ASSERT_EQ(NumTableFilesAtLevel(0), 1);
+  for (int i = 0; i < 3; i++) {
+    ASSERT_EQ("", Get("key" + ToString(i)));
+  }
+  for (int i = 0; i < 3; i++) {
+    // Each file contains a different key which will be dropped later.
+    ASSERT_OK(Put("a" + ToString(i), RandomString(&rnd, 500)));
+    ASSERT_OK(Delete("key" + ToString(i)));
+    ASSERT_OK(Put("z" + ToString(i), RandomString(&rnd, 500)));
+    Flush();
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  }
+  ASSERT_EQ(NumTableFilesAtLevel(0), 2);
+  for (int i = 0; i < 3; i++) {
+    ASSERT_EQ("NOT_FOUND", Get("key" + ToString(i)));
+  }
+}
+
+// Check that FIFO-with-TTL is not supported with max_open_files != -1.
+TEST_F(DBTest, FIFOCompactionWithTTLAndMaxOpenFilesTest) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.create_if_missing = true;
+  options.compaction_options_fifo.ttl = 600;  // seconds
+
+  // Check that it is not supported with max_open_files != -1.
+  options.max_open_files = 100;
+  options = CurrentOptions(options);
+  ASSERT_TRUE(TryReopen(options).IsNotSupported());
+
+  options.max_open_files = -1;
+  ASSERT_OK(TryReopen(options));
+}
+
+// Check that FIFO-with-TTL is supported only with BlockBasedTableFactory.
+TEST_F(DBTest, FIFOCompactionWithTTLAndVariousTableFormatsTest) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.create_if_missing = true;
+  options.compaction_options_fifo.ttl = 600;  // seconds
+
+  options = CurrentOptions(options);
+  options.table_factory.reset(NewBlockBasedTableFactory());
+  ASSERT_OK(TryReopen(options));
+
+  Destroy(options);
+  options.table_factory.reset(NewPlainTableFactory());
+  ASSERT_TRUE(TryReopen(options).IsNotSupported());
+
+  Destroy(options);
+  options.table_factory.reset(NewCuckooTableFactory());
+  ASSERT_TRUE(TryReopen(options).IsNotSupported());
+
+  Destroy(options);
+  options.table_factory.reset(NewAdaptiveTableFactory());
+  ASSERT_TRUE(TryReopen(options).IsNotSupported());
+}
+
+TEST_F(DBTest, DISABLED_FIFOCompactionWithTTLTest) {
+  Options options;
+  options.compaction_style = kCompactionStyleFIFO;
+  options.write_buffer_size = 10 << 10;  // 10KB
+  options.arena_block_size = 4096;
+  options.compression = kNoCompression;
+  options.create_if_missing = true;
+
+  // Test to make sure that all files with expired ttl are deleted on next
+  // manual compaction.
+  {
+    options.compaction_options_fifo.max_table_files_size = 150 << 10;  // 150KB
+    options.compaction_options_fifo.allow_compaction = false;
+    options.compaction_options_fifo.ttl = 600;  // seconds
+    options = CurrentOptions(options);
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    for (int i = 0; i < 10; i++) {
+      // Generate and flush a file about 10KB.
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+    // sleep for 5 seconds
+    env_->SleepForMicroseconds(5 * 1000 * 1000);
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+    // change ttl to 1 sec. So all files should be deleted on next compaction.
+    options.compaction_options_fifo.ttl = 1;
+    Reopen(options);
+
+    dbfull()->CompactRange(CompactRangeOptions(), nullptr, nullptr);
+    ASSERT_EQ(NumTableFilesAtLevel(0), 0);
+  }
+
+  // Test to make sure that all files with expired ttl are deleted on next
+  // automatic compaction.
+  {
+    options.compaction_options_fifo.max_table_files_size = 150 << 10;  // 150KB
+    options.compaction_options_fifo.allow_compaction = false;
+    options.compaction_options_fifo.ttl = 5;  // seconds
+    options = CurrentOptions(options);
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    for (int i = 0; i < 10; i++) {
+      // Generate and flush a file about 10KB.
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+    env_->SleepForMicroseconds(6 * 1000 * 1000);
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+    // Create 1 more file to trigger TTL compaction. The old files are dropped.
+    for (int i = 0; i < 1; i++) {
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    // Only the new 10 files remain.
+    ASSERT_EQ(NumTableFilesAtLevel(0), 1);
+    ASSERT_LE(SizeAtLevel(0),
+              options.compaction_options_fifo.max_table_files_size);
+  }
+
+  // Test that shows the fall back to size-based FIFO compaction if TTL-based
+  // deletion doesn't move the total size to be less than max_table_files_size.
+  {
+    options.write_buffer_size = 110 << 10;                             // 10KB
+    options.compaction_options_fifo.max_table_files_size = 150 << 10;  // 150KB
+    options.compaction_options_fifo.allow_compaction = false;
+    options.compaction_options_fifo.ttl = 5;  // seconds
+    options = CurrentOptions(options);
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    for (int i = 0; i < 3; i++) {
+      // Generate and flush a file about 10KB.
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 3);
+
+    env_->SleepForMicroseconds(6 * 1000 * 1000);
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 3);
+
+    for (int i = 0; i < 5; i++) {
+      for (int j = 0; j < 140; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    // Size limit is still guaranteed.
+    ASSERT_LE(SizeAtLevel(0),
+              options.compaction_options_fifo.max_table_files_size);
+  }
+
+  // Test with TTL + Intra-L0 compactions.
+  {
+    options.compaction_options_fifo.max_table_files_size = 150 << 10;  // 150KB
+    options.compaction_options_fifo.allow_compaction = true;
+    options.compaction_options_fifo.ttl = 5;  // seconds
+    options.level0_file_num_compaction_trigger = 6;
+    options = CurrentOptions(options);
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    for (int i = 0; i < 10; i++) {
+      // Generate and flush a file about 10KB.
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    // With Intra-L0 compaction, out of 10 files, 6 files will be compacted to 1
+    // (due to level0_file_num_compaction_trigger = 6).
+    // So total files = 1 + remaining 4 = 5.
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 5);
+
+    // Sleep for a little over ttl time.
+    env_->SleepForMicroseconds(6 * 1000 * 1000);
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 5);
+
+    // Create 10 more files. The old 5 files are dropped as their ttl expired.
+    for (int i = 0; i < 10; i++) {
+      for (int j = 0; j < 10; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    ASSERT_EQ(NumTableFilesAtLevel(0), 5);
+    ASSERT_LE(SizeAtLevel(0),
+              options.compaction_options_fifo.max_table_files_size);
+  }
+
+  // Test with large TTL + Intra-L0 compactions.
+  // Files dropped based on size, as ttl doesn't kick in.
+  {
+    options.write_buffer_size = 20 << 10;                               // 20K
+    options.compaction_options_fifo.max_table_files_size = 1500 << 10;  // 1.5MB
+    options.compaction_options_fifo.allow_compaction = true;
+    options.compaction_options_fifo.ttl = 60 * 60;  // 1 hour
+    options.level0_file_num_compaction_trigger = 6;
+    options = CurrentOptions(options);
+    DestroyAndReopen(options);
+
+    Random rnd(301);
+    for (int i = 0; i < 60; i++) {
+      // Generate and flush a file about 20KB.
+      for (int j = 0; j < 20; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+    // It should be compacted to 10 files.
+    ASSERT_EQ(NumTableFilesAtLevel(0), 10);
+
+    for (int i = 0; i < 60; i++) {
+      // Generate and flush a file about 20KB.
+      for (int j = 0; j < 20; j++) {
+        ASSERT_OK(Put(ToString(i * 20 + j + 2000), RandomString(&rnd, 980)));
+      }
+      Flush();
+    }
+    ASSERT_OK(dbfull()->TEST_WaitForCompact());
+
+    // It should be compacted to no more than 20 files.
+    ASSERT_GT(NumTableFilesAtLevel(0), 10);
+    ASSERT_LT(NumTableFilesAtLevel(0), 18);
+    // Size limit is still guaranteed.
+    ASSERT_LE(SizeAtLevel(0),
+              options.compaction_options_fifo.max_table_files_size);
+  }
+}
 #endif  // ROCKSDB_LITE
 
 #ifndef ROCKSDB_LITE

db/db_test2.cc

@@ -1028,6 +1028,7 @@ TEST_F(DBTest2, PresetCompressionDict) {
   const int kNumL0Files = 5;
 
   Options options;
+  options.env = CurrentOptions().env; // Make sure to use any custom env that the test is configured with.
   options.allow_concurrent_memtable_write = false;
   options.arena_block_size = kBlockSizeBytes;
   options.compaction_style = kCompactionStyleUniversal;
@@ -2234,6 +2235,71 @@ TEST_F(DBTest2, LowPriWrite) {
   Put("", "", wo);
   ASSERT_EQ(1, rate_limit_count.load());
 }
+
+TEST_F(DBTest2, RateLimitedCompactionReads) {
+  // compaction input has 512KB data
+  const int kNumKeysPerFile = 128;
+  const int kBytesPerKey = 1024;
+  const int kNumL0Files = 4;
+
+  for (auto use_direct_io : {false, true}) {
+    if (use_direct_io && !IsDirectIOSupported()) {
+      continue;
+    }
+    Options options = CurrentOptions();
+    options.compression = kNoCompression;
+    options.level0_file_num_compaction_trigger = kNumL0Files;
+    options.memtable_factory.reset(new SpecialSkipListFactory(kNumKeysPerFile));
+    options.new_table_reader_for_compaction_inputs = true;
+    // takes roughly one second, split into 100 x 10ms intervals. Each interval
+    // permits 5.12KB, which is smaller than the block size, so this test
+    // exercises the code for chunking reads.
+    options.rate_limiter.reset(NewGenericRateLimiter(
+        static_cast<int64_t>(kNumL0Files * kNumKeysPerFile *
+                             kBytesPerKey) /* rate_bytes_per_sec */,
+        10 * 1000 /* refill_period_us */, 10 /* fairness */,
+        RateLimiter::Mode::kReadsOnly));
+    options.use_direct_io_for_flush_and_compaction = use_direct_io;
+    BlockBasedTableOptions bbto;
+    bbto.block_size = 16384;
+    bbto.no_block_cache = true;
+    options.table_factory.reset(new BlockBasedTableFactory(bbto));
+    DestroyAndReopen(options);
+
+    for (int i = 0; i < kNumL0Files; ++i) {
+      for (int j = 0; j <= kNumKeysPerFile; ++j) {
+        ASSERT_OK(Put(Key(j), DummyString(kBytesPerKey)));
+      }
+      dbfull()->TEST_WaitForFlushMemTable();
+      ASSERT_EQ(i + 1, NumTableFilesAtLevel(0));
+    }
+    dbfull()->TEST_WaitForCompact();
+    ASSERT_EQ(0, NumTableFilesAtLevel(0));
+
+    ASSERT_EQ(0, options.rate_limiter->GetTotalBytesThrough(Env::IO_HIGH));
+    // should be slightly above 512KB due to non-data blocks read. Arbitrarily
+    // chose 1MB as the upper bound on the total bytes read.
+    size_t rate_limited_bytes =
+        options.rate_limiter->GetTotalBytesThrough(Env::IO_LOW);
+    ASSERT_GE(
+        rate_limited_bytes,
+        static_cast<size_t>(kNumKeysPerFile * kBytesPerKey * kNumL0Files));
+    ASSERT_LT(
+        rate_limited_bytes,
+        static_cast<size_t>(2 * kNumKeysPerFile * kBytesPerKey * kNumL0Files));
+
+    Iterator* iter = db_->NewIterator(ReadOptions());
+    for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
+      ASSERT_EQ(iter->value().ToString(), DummyString(kBytesPerKey));
+    }
+    delete iter;
+    // bytes read for user iterator shouldn't count against the rate limit.
+    ASSERT_EQ(rate_limited_bytes,
+              static_cast<size_t>(
+                  options.rate_limiter->GetTotalBytesThrough(Env::IO_LOW)));
+  }
+}
+
 }  // namespace rocksdb
 
 int main(int argc, char** argv) {

db/db_test_util.cc

@@ -9,6 +9,7 @@
 
 #include "db/db_test_util.h"
 #include "db/forward_iterator.h"
+#include "rocksdb/env_encryption.h"
 
 namespace rocksdb {
 
@@ -42,9 +43,12 @@ SpecialEnv::SpecialEnv(Env* base)
   table_write_callback_ = nullptr;
 }
 
+ROT13BlockCipher rot13Cipher_(16);
+
 DBTestBase::DBTestBase(const std::string path)
     : mem_env_(!getenv("MEM_ENV") ? nullptr : new MockEnv(Env::Default())),
-      env_(new SpecialEnv(mem_env_ ? mem_env_ : Env::Default())),
+      encrypted_env_(!getenv("ENCRYPTED_ENV") ? nullptr : NewEncryptedEnv(mem_env_ ? mem_env_ : Env::Default(), new CTREncryptionProvider(rot13Cipher_))),
+      env_(new SpecialEnv(encrypted_env_ ? encrypted_env_ : (mem_env_ ? mem_env_ : Env::Default()))),
       option_config_(kDefault) {
   env_->SetBackgroundThreads(1, Env::LOW);
   env_->SetBackgroundThreads(1, Env::HIGH);
@@ -281,6 +285,7 @@ Options DBTestBase::GetOptions(
       "NewWritableFile:O_DIRECT");
 #endif
 
+  bool can_allow_mmap = IsMemoryMappedAccessSupported();
   switch (option_config) {
 #ifndef ROCKSDB_LITE
     case kHashSkipList:
@@ -291,14 +296,14 @@ Options DBTestBase::GetOptions(
     case kPlainTableFirstBytePrefix:
       options.table_factory.reset(new PlainTableFactory());
       options.prefix_extractor.reset(NewFixedPrefixTransform(1));
-      options.allow_mmap_reads = true;
+      options.allow_mmap_reads = can_allow_mmap;
       options.max_sequential_skip_in_iterations = 999999;
       set_block_based_table_factory = false;
       break;
     case kPlainTableCappedPrefix:
       options.table_factory.reset(new PlainTableFactory());
       options.prefix_extractor.reset(NewCappedPrefixTransform(8));
-      options.allow_mmap_reads = true;
+      options.allow_mmap_reads = can_allow_mmap;
       options.max_sequential_skip_in_iterations = 999999;
       set_block_based_table_factory = false;
       break;
@@ -312,7 +317,7 @@ Options DBTestBase::GetOptions(
     case kPlainTableAllBytesPrefix:
       options.table_factory.reset(new PlainTableFactory());
       options.prefix_extractor.reset(NewNoopTransform());
-      options.allow_mmap_reads = true;
+      options.allow_mmap_reads = can_allow_mmap;
       options.max_sequential_skip_in_iterations = 999999;
       set_block_based_table_factory = false;
       break;
@@ -364,7 +369,7 @@ Options DBTestBase::GetOptions(
       options.wal_dir = alternative_wal_dir_;
       // mmap reads should be orthogonal to WalDir setting, so we piggyback to
       // this option config to test mmap reads as well
-      options.allow_mmap_reads = true;
+      options.allow_mmap_reads = can_allow_mmap;
       break;
     case kManifestFileSize:
       options.max_manifest_file_size = 50;  // 50 bytes
@@ -384,7 +389,7 @@ Options DBTestBase::GetOptions(
       options.num_levels = 8;
       break;
     case kCompressedBlockCache:
-      options.allow_mmap_writes = true;
+      options.allow_mmap_writes = can_allow_mmap;
       table_options.block_cache_compressed = NewLRUCache(8 * 1024 * 1024);
       break;
     case kInfiniteMaxOpenFiles:
@@ -469,6 +474,12 @@ Options DBTestBase::GetOptions(
       options.enable_pipelined_write = true;
       break;
     }
+    case kConcurrentWALWrites: {
+      // This options optimize 2PC commit path
+      options.concurrent_prepare = true;
+      options.manual_wal_flush = true;
+      break;
+    }
 
     default:
       break;
@@ -592,6 +603,10 @@ bool DBTestBase::IsDirectIOSupported() {
   return s.ok();
 }
 
+bool DBTestBase::IsMemoryMappedAccessSupported() const {
+  return (!encrypted_env_);
+}
+
 Status DBTestBase::Flush(int cf) {
   if (cf == 0) {
     return db_->Flush(FlushOptions());

db/db_test_util.h

@@ -247,6 +247,13 @@ class SpecialEnv : public EnvWrapper {
         }
       }
       Status Truncate(uint64_t size) override { return base_->Truncate(size); }
+      Status RangeSync(uint64_t offset, uint64_t nbytes) override {
+        Status s = base_->RangeSync(offset, nbytes);
+#if !(defined NDEBUG) || !defined(OS_WIN)
+        TEST_SYNC_POINT_CALLBACK("SpecialEnv::SStableFile::RangeSync", &s);
+#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+        return s;
+      }
       Status Close() override {
 // SyncPoint is not supported in Released Windows Mode.
 #if !(defined NDEBUG) || !defined(OS_WIN)
@@ -256,7 +263,11 @@ class SpecialEnv : public EnvWrapper {
         TEST_SYNC_POINT_CALLBACK("DBTestWritableFile.GetPreallocationStatus",
                                  &preallocation_size);
 #endif  // !(defined NDEBUG) || !defined(OS_WIN)
-        return base_->Close();
+        Status s = base_->Close();
+#if !(defined NDEBUG) || !defined(OS_WIN)
+        TEST_SYNC_POINT_CALLBACK("SpecialEnv::SStableFile::Close", &s);
+#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+        return s;
       }
       Status Flush() override { return base_->Flush(); }
       Status Sync() override {
@@ -264,7 +275,11 @@ class SpecialEnv : public EnvWrapper {
         while (env_->delay_sstable_sync_.load(std::memory_order_acquire)) {
           env_->SleepForMicroseconds(100000);
         }
-        return base_->Sync();
+        Status s = base_->Sync();
+#if !(defined NDEBUG) || !defined(OS_WIN)
+        TEST_SYNC_POINT_CALLBACK("SpecialEnv::SStableFile::Sync", &s);
+#endif  // !(defined NDEBUG) || !defined(OS_WIN)
+        return s;
       }
       void SetIOPriority(Env::IOPriority pri) override {
         base_->SetIOPriority(pri);
@@ -641,13 +656,14 @@ class DBTestBase : public testing::Test {
     kRecycleLogFiles = 28,
     kConcurrentSkipList = 29,
     kPipelinedWrite = 30,
-    kEnd = 31,
-    kDirectIO = 32,
-    kLevelSubcompactions = 33,
-    kUniversalSubcompactions = 34,
-    kBlockBasedTableWithIndexRestartInterval = 35,
-    kBlockBasedTableWithPartitionedIndex = 36,
-    kPartitionedFilterWithNewTableReaderForCompactions = 37,
+    kConcurrentWALWrites = 31,
+    kEnd = 32,
+    kDirectIO = 33,
+    kLevelSubcompactions = 34,
+    kUniversalSubcompactions = 35,
+    kBlockBasedTableWithIndexRestartInterval = 36,
+    kBlockBasedTableWithPartitionedIndex = 37,
+    kPartitionedFilterWithNewTableReaderForCompactions = 38,
   };
 
  public:
@@ -655,6 +671,7 @@ class DBTestBase : public testing::Test {
   std::string alternative_wal_dir_;
   std::string alternative_db_log_dir_;
   MockEnv* mem_env_;
+  Env* encrypted_env_;
   SpecialEnv* env_;
   DB* db_;
   std::vector<ColumnFamilyHandle*> handles_;
@@ -758,6 +775,8 @@ class DBTestBase : public testing::Test {
 
   bool IsDirectIOSupported();
 
+  bool IsMemoryMappedAccessSupported() const;
+
   Status Flush(int cf = 0);
 
   Status Put(const Slice& k, const Slice& v, WriteOptions wo = WriteOptions());

db/db_wal_test.cc

@@ -120,9 +120,11 @@ TEST_F(DBWALTest, SyncWALNotWaitWrite) {
   rocksdb::SyncPoint::GetInstance()->EnableProcessing();
 
   rocksdb::port::Thread thread([&]() { ASSERT_OK(Put("foo2", "bar2")); });
-  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
+  // Moving this to SyncWAL before the actual fsync
+  // TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
   ASSERT_OK(db_->SyncWAL());
-  TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
+  // Moving this to SyncWAL after actual fsync
+  // TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
 
   thread.join();
 
@@ -658,6 +660,7 @@ TEST_F(DBWALTest, PartOfWritesWithWALDisabled) {
   ASSERT_OK(Flush(0));
   ASSERT_OK(Put(0, "key", "v5", wal_on));  // seq id 5
   ASSERT_EQ("v5", Get(0, "key"));
+  dbfull()->FlushWAL(false);
   // Simulate a crash.
   fault_env->SetFilesystemActive(false);
   Close();
@@ -727,6 +730,7 @@ class RecoveryTestHelper {
         batch.Put(key, value);
         WriteBatchInternal::SetSequence(&batch, seq);
         current_log_writer->AddRecord(WriteBatchInternal::Contents(&batch));
+        versions->SetLastToBeWrittenSequence(seq);
         versions->SetLastSequence(seq);
       }
     }
@@ -1111,6 +1115,7 @@ TEST_F(DBWALTest, RecoverWithoutFlushMultipleCF) {
   ASSERT_EQ(3, countWalFiles());
   Flush(1);
   ASSERT_OK(Put(2, "key7", kLargeValue));
+  dbfull()->FlushWAL(false);
   ASSERT_EQ(4, countWalFiles());
 
   // Reopen twice and validate.

db/db_write_test.cc

@@ -69,6 +69,7 @@ TEST_P(DBWriteTest, ReturnSeuqneceNumberMultiThreaded) {
 
 INSTANTIATE_TEST_CASE_P(DBWriteTestInstance, DBWriteTest,
                         testing::Values(DBTestBase::kDefault,
+                                        DBTestBase::kConcurrentWALWrites,
                                         DBTestBase::kPipelinedWrite));
 
 }  // namespace rocksdb

db/dbformat.h

@@ -78,7 +78,9 @@ struct ParsedInternalKey {
   SequenceNumber sequence;
   ValueType type;
 
-  ParsedInternalKey() { }  // Intentionally left uninitialized (for speed)
+  ParsedInternalKey()
+      : sequence(kMaxSequenceNumber)  // Make code analyzer happy
+  {}  // Intentionally left uninitialized (for speed)
   ParsedInternalKey(const Slice& u, const SequenceNumber& seq, ValueType t)
       : user_key(u), sequence(seq), type(t) { }
   std::string DebugString(bool hex = false) const;

db/event_helpers.cc

@@ -37,6 +37,24 @@ void EventHelpers::NotifyTableFileCreationStarted(
 }
 #endif  // !ROCKSDB_LITE
 
+void EventHelpers::NotifyOnBackgroundError(
+    const std::vector<std::shared_ptr<EventListener>>& listeners,
+    BackgroundErrorReason reason, Status* bg_error,
+    InstrumentedMutex* db_mutex) {
+#ifndef ROCKSDB_LITE
+  if (listeners.size() == 0U) {
+    return;
+  }
+  db_mutex->AssertHeld();
+  // release lock while notifying events
+  db_mutex->Unlock();
+  for (auto& listener : listeners) {
+    listener->OnBackgroundError(reason, bg_error);
+  }
+  db_mutex->Lock();
+#endif  // ROCKSDB_LITE
+}
+
 void EventHelpers::LogAndNotifyTableFileCreationFinished(
     EventLogger* event_logger,
     const std::vector<std::shared_ptr<EventListener>>& listeners,

db/event_helpers.h

@@ -25,6 +25,10 @@ class EventHelpers {
       const std::string& db_name, const std::string& cf_name,
       const std::string& file_path, int job_id, TableFileCreationReason reason);
 #endif  // !ROCKSDB_LITE
+  static void NotifyOnBackgroundError(
+      const std::vector<std::shared_ptr<EventListener>>& listeners,
+      BackgroundErrorReason reason, Status* bg_error,
+      InstrumentedMutex* db_mutex);
   static void LogAndNotifyTableFileCreationFinished(
       EventLogger* event_logger,
       const std::vector<std::shared_ptr<EventListener>>& listeners,

db/external_sst_file_ingestion_job.cc

@@ -144,6 +144,8 @@ Status ExternalSstFileIngestionJob::NeedsFlush(bool* flush_needed) {
   return status;
 }
 
+// REQUIRES: we have become the only writer by entering both write_thread_ and
+// nonmem_write_thread_
 Status ExternalSstFileIngestionJob::Run() {
   Status status;
 #ifndef NDEBUG
@@ -162,6 +164,8 @@ Status ExternalSstFileIngestionJob::Run() {
     // if the dont overlap with any ranges since we have snapshots
     force_global_seqno = true;
   }
+  // It is safe to use this instead of LastToBeWrittenSequence since we are
+  // the only active writer, and hence they are equal
   const SequenceNumber last_seqno = versions_->LastSequence();
   SuperVersion* super_version = cfd_->GetSuperVersion();
   edit_.SetColumnFamily(cfd_->GetID());
@@ -195,6 +199,7 @@ Status ExternalSstFileIngestionJob::Run() {
   }
 
   if (consumed_seqno) {
+    versions_->SetLastToBeWrittenSequence(last_seqno + 1);
     versions_->SetLastSequence(last_seqno + 1);
   }
 
@@ -279,7 +284,8 @@ Status ExternalSstFileIngestionJob::GetIngestedFileInfo(
   if (!status.ok()) {
     return status;
   }
-  sst_file_reader.reset(new RandomAccessFileReader(std::move(sst_file)));
+  sst_file_reader.reset(new RandomAccessFileReader(std::move(sst_file),
+                                                   external_file));
 
   status = cfd_->ioptions()->table_factory->NewTableReader(
       TableReaderOptions(*cfd_->ioptions(), env_options_,

db/fault_injection_test.cc

@@ -410,6 +410,7 @@ TEST_P(FaultInjectionTest, WriteOptionSyncTest) {
   write_options.sync = true;
   ASSERT_OK(
       db_->Put(write_options, Key(2, &key_space), Value(2, &value_space)));
+  db_->FlushWAL(false);
 
   env_->SetFilesystemActive(false);
   NoWriteTestReopenWithFault(kResetDropAndDeleteUnsynced);
@@ -494,7 +495,7 @@ TEST_P(FaultInjectionTest, ManualLogSyncTest) {
   ASSERT_OK(db_->Flush(flush_options));
   ASSERT_OK(
       db_->Put(write_options, Key(2, &key_space), Value(2, &value_space)));
-  ASSERT_OK(db_->SyncWAL());
+  ASSERT_OK(db_->FlushWAL(true));
 
   env_->SetFilesystemActive(false);
   NoWriteTestReopenWithFault(kResetDropAndDeleteUnsynced);

db/flush_job.cc

@@ -296,6 +296,11 @@ Status FlushJob::WriteLevel0Table() {
                                &output_compression_);
       EnvOptions optimized_env_options =
           db_options_.env->OptimizeForCompactionTableWrite(env_options_, db_options_);
+
+      int64_t _current_time = 0;
+      db_options_.env->GetCurrentTime(&_current_time);  // ignore error
+      const uint64_t current_time = static_cast<uint64_t>(_current_time);
+
       s = BuildTable(
           dbname_, db_options_.env, *cfd_->ioptions(), mutable_cf_options_,
           optimized_env_options, cfd_->table_cache(), iter.get(),
@@ -306,7 +311,7 @@ Status FlushJob::WriteLevel0Table() {
           cfd_->ioptions()->compression_opts,
           mutable_cf_options_.paranoid_file_checks, cfd_->internal_stats(),
           TableFileCreationReason::kFlush, event_logger_, job_context_->job_id,
-          Env::IO_HIGH, &table_properties_, 0 /* level */);
+          Env::IO_HIGH, &table_properties_, 0 /* level */, current_time);
       LogFlush(db_options_.info_log);
     }
     ROCKS_LOG_INFO(db_options_.info_log,

db/flush_job.h

@@ -68,7 +68,7 @@ class FlushJob {
   ~FlushJob();
 
   // Require db_mutex held.
-  // Once PickMemTable() is called, either Run() or Cancel() has to be call.
+  // Once PickMemTable() is called, either Run() or Cancel() has to be called.
   void PickMemTable();
   Status Run(FileMetaData* file_meta = nullptr);
   void Cancel();

db/internal_stats.cc

@@ -542,7 +542,7 @@ bool InternalStats::HandleDBStats(std::string* value, Slice suffix) {
 
 bool InternalStats::HandleSsTables(std::string* value, Slice suffix) {
   auto* current = cfd_->current();
-  *value = current->DebugString();
+  *value = current->DebugString(true, true);
   return true;
 }
 

db/internal_stats.h

@@ -270,10 +270,15 @@ class InternalStats {
     ++cf_stats_count_[type];
   }
 
-  void AddDBStats(InternalDBStatsType type, uint64_t value) {
+  void AddDBStats(InternalDBStatsType type, uint64_t value,
+                  bool concurrent = false) {
     auto& v = db_stats_[type];
-    v.store(v.load(std::memory_order_relaxed) + value,
-            std::memory_order_relaxed);
+    if (concurrent) {
+      v.fetch_add(value, std::memory_order_relaxed);
+    } else {
+      v.store(v.load(std::memory_order_relaxed) + value,
+              std::memory_order_relaxed);
+    }
   }
 
   uint64_t GetDBStats(InternalDBStatsType type) {
@@ -548,7 +553,8 @@ class InternalStats {
 
   void AddCFStats(InternalCFStatsType type, uint64_t value) {}
 
-  void AddDBStats(InternalDBStatsType type, uint64_t value) {}
+  void AddDBStats(InternalDBStatsType type, uint64_t value,
+                  bool concurrent = false) {}
 
   HistogramImpl* GetFileReadHist(int level) { return nullptr; }
 

db/listener_test.cc

@@ -112,6 +112,7 @@ TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
   const int kNumL0Files = 4;
 
   Options options;
+  options.env = CurrentOptions().env;
   options.create_if_missing = true;
   options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
   options.compaction_style = kCompactionStyleLevel;
@@ -231,6 +232,7 @@ class TestFlushListener : public EventListener {
 
 TEST_F(EventListenerTest, OnSingleDBFlushTest) {
   Options options;
+  options.env = CurrentOptions().env;
   options.write_buffer_size = k110KB;
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
@@ -267,6 +269,7 @@ TEST_F(EventListenerTest, OnSingleDBFlushTest) {
 
 TEST_F(EventListenerTest, MultiCF) {
   Options options;
+  options.env = CurrentOptions().env;
   options.write_buffer_size = k110KB;
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
@@ -302,6 +305,7 @@ TEST_F(EventListenerTest, MultiCF) {
 
 TEST_F(EventListenerTest, MultiDBMultiListeners) {
   Options options;
+  options.env = CurrentOptions().env;
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
@@ -384,6 +388,7 @@ TEST_F(EventListenerTest, MultiDBMultiListeners) {
 
 TEST_F(EventListenerTest, DisableBGCompaction) {
   Options options;
+  options.env = CurrentOptions().env;
 #ifdef ROCKSDB_USING_THREAD_STATUS
   options.enable_thread_tracking = true;
 #endif  // ROCKSDB_USING_THREAD_STATUS
@@ -431,6 +436,7 @@ class TestCompactionReasonListener : public EventListener {
 
 TEST_F(EventListenerTest, CompactionReasonLevel) {
   Options options;
+  options.env = CurrentOptions().env;
   options.create_if_missing = true;
   options.memtable_factory.reset(
       new SpecialSkipListFactory(DBTestBase::kNumKeysByGenerateNewRandomFile));
@@ -496,6 +502,7 @@ TEST_F(EventListenerTest, CompactionReasonLevel) {
 
 TEST_F(EventListenerTest, CompactionReasonUniversal) {
   Options options;
+  options.env = CurrentOptions().env;
   options.create_if_missing = true;
   options.memtable_factory.reset(
       new SpecialSkipListFactory(DBTestBase::kNumKeysByGenerateNewRandomFile));
@@ -557,6 +564,7 @@ TEST_F(EventListenerTest, CompactionReasonUniversal) {
 
 TEST_F(EventListenerTest, CompactionReasonFIFO) {
   Options options;
+  options.env = CurrentOptions().env;
   options.create_if_missing = true;
   options.memtable_factory.reset(
       new SpecialSkipListFactory(DBTestBase::kNumKeysByGenerateNewRandomFile));
@@ -779,6 +787,7 @@ TEST_F(EventListenerTest, ColumnFamilyHandleDeletionStartedListenerTest) {
   auto listener =
       std::make_shared<ColumnFamilyHandleDeletionStartedListener>(cfs);
   Options options;
+  options.env = CurrentOptions().env;
   options.create_if_missing = true;
   options.listeners.push_back(listener);
   CreateAndReopenWithCF(cfs, options);
@@ -790,6 +799,92 @@ TEST_F(EventListenerTest, ColumnFamilyHandleDeletionStartedListenerTest) {
   ASSERT_EQ(listener->getCounter(), 3);
 }
 
+class BackgroundErrorListener : public EventListener {
+ private:
+  SpecialEnv* env_;
+  int counter_;
+
+ public:
+  BackgroundErrorListener(SpecialEnv* env) : env_(env), counter_(0) {}
+
+  void OnBackgroundError(BackgroundErrorReason reason, Status* bg_error) override {
+    if (counter_ == 0) {
+      // suppress the first error and disable write-dropping such that a retry
+      // can succeed.
+      *bg_error = Status::OK();
+      env_->drop_writes_.store(false, std::memory_order_release);
+      env_->no_slowdown_ = false;
+    }
+    ++counter_;
+  }
+
+  int counter() { return counter_; }
+};
+
+TEST_F(EventListenerTest, BackgroundErrorListenerFailedFlushTest) {
+  auto listener = std::make_shared<BackgroundErrorListener>(env_);
+  Options options;
+  options.create_if_missing = true;
+  options.env = env_;
+  options.listeners.push_back(listener);
+  options.memtable_factory.reset(new SpecialSkipListFactory(1));
+  options.paranoid_checks = true;
+  DestroyAndReopen(options);
+
+  // the usual TEST_WaitForFlushMemTable() doesn't work for failed flushes, so
+  // forge a custom one for the failed flush case.
+  rocksdb::SyncPoint::GetInstance()->LoadDependency(
+      {{"DBImpl::BGWorkFlush:done",
+        "EventListenerTest:BackgroundErrorListenerFailedFlushTest:1"}});
+  rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+  env_->drop_writes_.store(true, std::memory_order_release);
+  env_->no_slowdown_ = true;
+
+  ASSERT_OK(Put("key0", "val"));
+  ASSERT_OK(Put("key1", "val"));
+  TEST_SYNC_POINT("EventListenerTest:BackgroundErrorListenerFailedFlushTest:1");
+  ASSERT_EQ(1, listener->counter());
+  ASSERT_OK(Put("key2", "val"));
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
+  ASSERT_EQ(1, NumTableFilesAtLevel(0));
+}
+
+TEST_F(EventListenerTest, BackgroundErrorListenerFailedCompactionTest) {
+  auto listener = std::make_shared<BackgroundErrorListener>(env_);
+  Options options;
+  options.create_if_missing = true;
+  options.disable_auto_compactions = true;
+  options.env = env_;
+  options.level0_file_num_compaction_trigger = 2;
+  options.listeners.push_back(listener);
+  options.memtable_factory.reset(new SpecialSkipListFactory(2));
+  options.paranoid_checks = true;
+  DestroyAndReopen(options);
+
+  // third iteration triggers the second memtable's flush
+  for (int i = 0; i < 3; ++i) {
+    ASSERT_OK(Put("key0", "val"));
+    if (i > 0) {
+      ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
+    }
+    ASSERT_OK(Put("key1", "val"));
+  }
+  ASSERT_EQ(2, NumTableFilesAtLevel(0));
+
+  env_->drop_writes_.store(true, std::memory_order_release);
+  env_->no_slowdown_ = true;
+  ASSERT_OK(dbfull()->SetOptions({{"disable_auto_compactions", "false"}}));
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  ASSERT_EQ(1, listener->counter());
+
+  // trigger flush so compaction is triggered again; this time it succeeds
+  ASSERT_OK(Put("key0", "val"));
+  ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
+  ASSERT_OK(dbfull()->TEST_WaitForCompact());
+  ASSERT_EQ(0, NumTableFilesAtLevel(0));
+}
+
 }  // namespace rocksdb
 
 #endif  // ROCKSDB_LITE

db/log_writer.cc

@@ -18,20 +18,22 @@
 namespace rocksdb {
 namespace log {
 
-Writer::Writer(unique_ptr<WritableFileWriter>&& dest,
-               uint64_t log_number, bool recycle_log_files)
+Writer::Writer(unique_ptr<WritableFileWriter>&& dest, uint64_t log_number,
+               bool recycle_log_files, bool manual_flush)
     : dest_(std::move(dest)),
       block_offset_(0),
       log_number_(log_number),
-      recycle_log_files_(recycle_log_files) {
+      recycle_log_files_(recycle_log_files),
+      manual_flush_(manual_flush) {
   for (int i = 0; i <= kMaxRecordType; i++) {
     char t = static_cast<char>(i);
     type_crc_[i] = crc32c::Value(&t, 1);
   }
 }
 
-Writer::~Writer() {
-}
+Writer::~Writer() { WriteBuffer(); }
+
+Status Writer::WriteBuffer() { return dest_->Flush(); }
 
 Status Writer::AddRecord(const Slice& slice) {
   const char* ptr = slice.data();
@@ -127,7 +129,9 @@ Status Writer::EmitPhysicalRecord(RecordType t, const char* ptr, size_t n) {
   if (s.ok()) {
     s = dest_->Append(Slice(ptr, n));
     if (s.ok()) {
-      s = dest_->Flush();
+      if (!manual_flush_) {
+        s = dest_->Flush();
+      }
     }
   }
   block_offset_ += header_size + n;

db/log_writer.h

@@ -72,8 +72,8 @@ class Writer {
   // Create a writer that will append data to "*dest".
   // "*dest" must be initially empty.
   // "*dest" must remain live while this Writer is in use.
-  explicit Writer(unique_ptr<WritableFileWriter>&& dest,
-                  uint64_t log_number, bool recycle_log_files);
+  explicit Writer(unique_ptr<WritableFileWriter>&& dest, uint64_t log_number,
+                  bool recycle_log_files, bool manual_flush = false);
   ~Writer();
 
   Status AddRecord(const Slice& slice);
@@ -83,6 +83,8 @@ class Writer {
 
   uint64_t get_log_number() const { return log_number_; }
 
+  Status WriteBuffer();
+
  private:
   unique_ptr<WritableFileWriter> dest_;
   size_t block_offset_;       // Current offset in block
@@ -96,6 +98,10 @@ class Writer {
 
   Status EmitPhysicalRecord(RecordType type, const char* ptr, size_t length);
 
+  // If true, it does not flush after each write. Instead it relies on the upper
+  // layer to manually does the flush by calling ::WriteBuffer()
+  bool manual_flush_;
+
   // No copying allowed
   Writer(const Writer&);
   void operator=(const Writer&);

db/memtable.cc

@@ -643,6 +643,14 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
   }
   PERF_TIMER_GUARD(get_from_memtable_time);
 
+  std::unique_ptr<InternalIterator> range_del_iter(
+      NewRangeTombstoneIterator(read_opts));
+  Status status = range_del_agg->AddTombstones(std::move(range_del_iter));
+  if (!status.ok()) {
+    *s = status;
+    return false;
+  }
+
   Slice user_key = key.user_key();
   bool found_final_value = false;
   bool merge_in_progress = s->IsMergeInProgress();
@@ -658,13 +666,6 @@ bool MemTable::Get(const LookupKey& key, std::string* value, Status* s,
     if (prefix_bloom_) {
       PERF_COUNTER_ADD(bloom_memtable_hit_count, 1);
     }
-    std::unique_ptr<InternalIterator> range_del_iter(
-        NewRangeTombstoneIterator(read_opts));
-    Status status = range_del_agg->AddTombstones(std::move(range_del_iter));
-    if (!status.ok()) {
-      *s = status;
-      return false;
-    }
     Saver saver;
     saver.status = s;
     saver.found_final_value = &found_final_value;

db/range_del_aggregator.cc

@@ -412,8 +412,8 @@ void RangeDelAggregator::AddToBuilder(
 
   // Note the order in which tombstones are stored is insignificant since we
   // insert them into a std::map on the read path.
-  bool first_added = false;
   while (stripe_map_iter != rep_->stripe_map_.end()) {
+    bool first_added = false;
     for (auto tombstone_map_iter = stripe_map_iter->second.raw_map.begin();
          tombstone_map_iter != stripe_map_iter->second.raw_map.end();
          ++tombstone_map_iter) {
@@ -452,7 +452,7 @@ void RangeDelAggregator::AddToBuilder(
       builder->Add(ikey_and_end_key.first.Encode(), ikey_and_end_key.second);
       if (!first_added) {
         first_added = true;
-        InternalKey smallest_candidate = std::move(ikey_and_end_key.first);;
+        InternalKey smallest_candidate = std::move(ikey_and_end_key.first);
         if (lower_bound != nullptr &&
             icmp_.user_comparator()->Compare(smallest_candidate.user_key(),
                                              *lower_bound) <= 0) {

db/repair.cc

@@ -380,6 +380,11 @@ class Repairer {
       ScopedArenaIterator iter(mem->NewIterator(ro, &arena));
       EnvOptions optimized_env_options =
           env_->OptimizeForCompactionTableWrite(env_options_, immutable_db_options_);
+
+      int64_t _current_time = 0;
+      status = env_->GetCurrentTime(&_current_time);  // ignore error
+      const uint64_t current_time = static_cast<uint64_t>(_current_time);
+
       status = BuildTable(
           dbname_, env_, *cfd->ioptions(), *cfd->GetLatestMutableCFOptions(),
           optimized_env_options, table_cache_, iter.get(),
@@ -387,7 +392,9 @@ class Repairer {
           &meta, cfd->internal_comparator(),
           cfd->int_tbl_prop_collector_factories(), cfd->GetID(), cfd->GetName(),
           {}, kMaxSequenceNumber, kNoCompression, CompressionOptions(), false,
-          nullptr /* internal_stats */, TableFileCreationReason::kRecovery);
+          nullptr /* internal_stats */, TableFileCreationReason::kRecovery,
+          nullptr /* event_logger */, 0 /* job_id */, Env::IO_HIGH,
+          nullptr /* table_properties */, -1 /* level */, current_time);
       ROCKS_LOG_INFO(db_options_.info_log,
                      "Log #%" PRIu64 ": %d ops saved to Table #%" PRIu64 " %s",
                      log, counter, meta.fd.GetNumber(),
@@ -522,6 +529,7 @@ class Repairer {
         max_sequence = tables_[i].max_sequence;
       }
     }
+    vset_.SetLastToBeWrittenSequence(max_sequence);
     vset_.SetLastSequence(max_sequence);
 
     for (const auto& cf_id_and_tables : cf_id_to_tables) {

db/table_cache.cc

@@ -89,7 +89,8 @@ Status TableCache::GetTableReader(
     const InternalKeyComparator& internal_comparator, const FileDescriptor& fd,
     bool sequential_mode, size_t readahead, bool record_read_stats,
     HistogramImpl* file_read_hist, unique_ptr<TableReader>* table_reader,
-    bool skip_filters, int level, bool prefetch_index_and_filter_in_cache) {
+    bool skip_filters, int level, bool prefetch_index_and_filter_in_cache,
+    bool for_compaction) {
   std::string fname =
       TableFileName(ioptions_.db_paths, fd.GetNumber(), fd.GetPathId());
   unique_ptr<RandomAccessFile> file;
@@ -105,9 +106,10 @@ Status TableCache::GetTableReader(
     }
     StopWatch sw(ioptions_.env, ioptions_.statistics, TABLE_OPEN_IO_MICROS);
     std::unique_ptr<RandomAccessFileReader> file_reader(
-        new RandomAccessFileReader(std::move(file), ioptions_.env,
+        new RandomAccessFileReader(std::move(file), fname, ioptions_.env,
                                    ioptions_.statistics, record_read_stats,
-                                   file_read_hist));
+                                   file_read_hist, ioptions_.rate_limiter,
+                                   for_compaction));
     s = ioptions_.table_factory->NewTableReader(
         TableReaderOptions(ioptions_, env_options, internal_comparator,
                            skip_filters, level),
@@ -203,7 +205,8 @@ InternalIterator* TableCache::NewIterator(
       s = GetTableReader(
           env_options, icomparator, fd, true /* sequential_mode */, readahead,
           !for_compaction /* record stats */, nullptr, &table_reader_unique_ptr,
-          false /* skip_filters */, level);
+          false /* skip_filters */, level,
+          true /* prefetch_index_and_filter_in_cache */, for_compaction);
       if (s.ok()) {
         table_reader = table_reader_unique_ptr.release();
       }

db/table_cache.h

@@ -134,7 +134,8 @@ class TableCache {
                         HistogramImpl* file_read_hist,
                         unique_ptr<TableReader>* table_reader,
                         bool skip_filters = false, int level = -1,
-                        bool prefetch_index_and_filter_in_cache = true);
+                        bool prefetch_index_and_filter_in_cache = true,
+                        bool for_compaction = false);
 
   const ImmutableCFOptions& ioptions_;
   const EnvOptions& env_options_;

db/version_edit.h

@@ -61,18 +61,30 @@ struct FileDescriptor {
   uint64_t GetFileSize() const { return file_size; }
 };
 
+struct FileSampledStats {
+  FileSampledStats() : num_reads_sampled(0) {}
+  FileSampledStats(const FileSampledStats& other) { *this = other; }
+  FileSampledStats& operator=(const FileSampledStats& other) {
+    num_reads_sampled = other.num_reads_sampled.load();
+    return *this;
+  }
+
+  // number of user reads to this file.
+  mutable std::atomic<uint64_t> num_reads_sampled;
+};
+
 struct FileMetaData {
-  int refs;
   FileDescriptor fd;
   InternalKey smallest;            // Smallest internal key served by table
   InternalKey largest;             // Largest internal key served by table
-  bool being_compacted;            // Is this file undergoing compaction?
   SequenceNumber smallest_seqno;   // The smallest seqno in this file
   SequenceNumber largest_seqno;    // The largest seqno in this file
 
   // Needs to be disposed when refs becomes 0.
   Cache::Handle* table_reader_handle;
 
+  FileSampledStats stats;
+
   // Stats for compensating deletion entries during compaction
 
   // File size compensated by deletion entry.
@@ -85,6 +97,10 @@ struct FileMetaData {
   uint64_t num_deletions;          // the number of deletion entries.
   uint64_t raw_key_size;           // total uncompressed key size.
   uint64_t raw_value_size;         // total uncompressed value size.
+
+  int refs;  // Reference count
+
+  bool being_compacted;        // Is this file undergoing compaction?
   bool init_stats_from_file;   // true if the data-entry stats of this file
                                // has initialized from file.
 
@@ -92,9 +108,7 @@ struct FileMetaData {
                                // file.
 
   FileMetaData()
-      : refs(0),
-        being_compacted(false),
-        smallest_seqno(kMaxSequenceNumber),
+      : smallest_seqno(kMaxSequenceNumber),
         largest_seqno(0),
         table_reader_handle(nullptr),
         compensated_file_size(0),
@@ -102,6 +116,8 @@ struct FileMetaData {
         num_deletions(0),
         raw_key_size(0),
         raw_value_size(0),
+        refs(0),
+        being_compacted(false),
         init_stats_from_file(false),
         marked_for_compaction(false) {}
 
@@ -117,10 +133,12 @@ struct FileMetaData {
   }
 };
 
-// A compressed copy of file meta data that just contain
-// smallest and largest key's slice
+// A compressed copy of file meta data that just contain minimum data needed
+// to server read operations, while still keeping the pointer to full metadata
+// of the file in case it is needed.
 struct FdWithKeyRange {
   FileDescriptor fd;
+  FileMetaData* file_metadata;  // Point to all metadata
   Slice smallest_key;    // slice that contain smallest key
   Slice largest_key;     // slice that contain largest key
 
@@ -130,8 +148,12 @@ struct FdWithKeyRange {
         largest_key() {
   }
 
-  FdWithKeyRange(FileDescriptor _fd, Slice _smallest_key, Slice _largest_key)
-      : fd(_fd), smallest_key(_smallest_key), largest_key(_largest_key) {}
+  FdWithKeyRange(FileDescriptor _fd, Slice _smallest_key, Slice _largest_key,
+                 FileMetaData* _file_metadata)
+      : fd(_fd),
+        file_metadata(_file_metadata),
+        smallest_key(_smallest_key),
+        largest_key(_largest_key) {}
 };
 
 // Data structure to store an array of FdWithKeyRange in one level

db/version_set.cc

@@ -32,6 +32,7 @@
 #include "db/pinned_iterators_manager.h"
 #include "db/table_cache.h"
 #include "db/version_builder.h"
+#include "monitoring/file_read_sample.h"
 #include "monitoring/perf_context_imp.h"
 #include "rocksdb/env.h"
 #include "rocksdb/merge_operator.h"
@@ -120,7 +121,7 @@ class FilePicker {
     }
   }
 
-  int GetCurrentLevel() { return returned_file_level_; }
+  int GetCurrentLevel() const { return curr_level_; }
 
   FdWithKeyRange* GetNextFile() {
     while (!search_ended_) {  // Loops over different levels.
@@ -368,6 +369,7 @@ void DoGenerateLevelFilesBrief(LevelFilesBrief* file_level,
 
     FdWithKeyRange& f = file_level->files[i];
     f.fd = files[i]->fd;
+    f.file_metadata = files[i];
     f.smallest_key = Slice(mem, smallest_size);
     f.largest_key = Slice(mem + smallest_size, largest_size);
   }
@@ -435,12 +437,12 @@ namespace {
 class LevelFileNumIterator : public InternalIterator {
  public:
   LevelFileNumIterator(const InternalKeyComparator& icmp,
-                       const LevelFilesBrief* flevel)
+                       const LevelFilesBrief* flevel, bool should_sample)
       : icmp_(icmp),
         flevel_(flevel),
         index_(static_cast<uint32_t>(flevel->num_files)),
-        current_value_(0, 0, 0) {  // Marks as invalid
-  }
+        current_value_(0, 0, 0),  // Marks as invalid
+        should_sample_(should_sample) {}
   virtual bool Valid() const override { return index_ < flevel_->num_files; }
   virtual void Seek(const Slice& target) override {
     index_ = FindFile(icmp_, *flevel_, target);
@@ -475,6 +477,9 @@ class LevelFileNumIterator : public InternalIterator {
     assert(Valid());
 
     auto file_meta = flevel_->files[index_];
+    if (should_sample_) {
+      sample_file_read_inc(file_meta.file_metadata);
+    }
     current_value_ = file_meta.fd;
     return Slice(reinterpret_cast<const char*>(&current_value_),
                  sizeof(FileDescriptor));
@@ -486,6 +491,7 @@ class LevelFileNumIterator : public InternalIterator {
   const LevelFilesBrief* flevel_;
   uint32_t index_;
   mutable FileDescriptor current_value_;
+  bool should_sample_;
 };
 
 class LevelFileIteratorState : public TwoLevelIteratorState {
@@ -590,15 +596,15 @@ Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
   // 2. Table is not present in table cache, we'll read the table properties
   // directly from the properties block in the file.
   std::unique_ptr<RandomAccessFile> file;
+  std::string file_name;
   if (fname != nullptr) {
-    s = ioptions->env->NewRandomAccessFile(
-        *fname, &file, vset_->env_options_);
+    file_name = *fname;
   } else {
-    s = ioptions->env->NewRandomAccessFile(
-        TableFileName(vset_->db_options_->db_paths, file_meta->fd.GetNumber(),
-                      file_meta->fd.GetPathId()),
-        &file, vset_->env_options_);
+    file_name =
+      TableFileName(vset_->db_options_->db_paths, file_meta->fd.GetNumber(),
+                    file_meta->fd.GetPathId());
   }
+  s = ioptions->env->NewRandomAccessFile(file_name, &file, vset_->env_options_);
   if (!s.ok()) {
     return s;
   }
@@ -607,7 +613,7 @@ Status Version::GetTableProperties(std::shared_ptr<const TableProperties>* tp,
   // By setting the magic number to kInvalidTableMagicNumber, we can by
   // pass the magic number check in the footer.
   std::unique_ptr<RandomAccessFileReader> file_reader(
-      new RandomAccessFileReader(std::move(file)));
+      new RandomAccessFileReader(std::move(file), file_name));
   s = ReadTableProperties(
       file_reader.get(), file_meta->fd.GetFileSize(),
       Footer::kInvalidTableMagicNumber /* table's magic number */, *ioptions, &raw_table_properties);
@@ -743,13 +749,11 @@ void Version::GetColumnFamilyMetaData(ColumnFamilyMetaData* cf_meta) {
         file_path = ioptions->db_paths.back().path;
       }
       files.emplace_back(
-          MakeTableFileName("", file->fd.GetNumber()),
-          file_path,
-          file->fd.GetFileSize(),
-          file->smallest_seqno,
-          file->largest_seqno,
+          MakeTableFileName("", file->fd.GetNumber()), file_path,
+          file->fd.GetFileSize(), file->smallest_seqno, file->largest_seqno,
           file->smallest.user_key().ToString(),
           file->largest.user_key().ToString(),
+          file->stats.num_reads_sampled.load(std::memory_order_relaxed),
           file->being_compacted);
       level_size += file->fd.GetFileSize();
     }
@@ -833,6 +837,8 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
     return;
   }
 
+  bool should_sample = should_sample_file_read();
+
   auto* arena = merge_iter_builder->GetArena();
   if (level == 0) {
     // Merge all level zero files together since they may overlap
@@ -843,6 +849,15 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
           range_del_agg, nullptr, cfd_->internal_stats()->GetFileReadHist(0),
           false, arena, false /* skip_filters */, 0 /* level */));
     }
+    if (should_sample) {
+      // Count ones for every L0 files. This is done per iterator creation
+      // rather than Seek(), while files in other levels are recored per seek.
+      // If users execute one range query per iterator, there may be some
+      // discrepancy here.
+      for (FileMetaData* meta : storage_info_.LevelFiles(0)) {
+        sample_file_read_inc(meta);
+      }
+    }
   } else {
     // For levels > 0, we can use a concatenating iterator that sequentially
     // walks through the non-overlapping files in the level, opening them
@@ -857,7 +872,8 @@ void Version::AddIteratorsForLevel(const ReadOptions& read_options,
                                IsFilterSkipped(level), level, range_del_agg);
     mem = arena->AllocateAligned(sizeof(LevelFileNumIterator));
     auto* first_level_iter = new (mem) LevelFileNumIterator(
-        cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level));
+        cfd_->internal_comparator(), &storage_info_.LevelFilesBrief(level),
+        should_sample_file_read());
     merge_iter_builder->AddIterator(
         NewTwoLevelIterator(state, first_level_iter, arena, false));
   }
@@ -982,6 +998,9 @@ void Version::Get(const ReadOptions& read_options, const LookupKey& k,
       user_comparator(), internal_comparator());
   FdWithKeyRange* f = fp.GetNextFile();
   while (f != nullptr) {
+    if (get_context.sample()) {
+      sample_file_read_inc(f->file_metadata);
+    }
     *status = table_cache_->Get(
         read_options, *internal_comparator(), f->fd, ikey, &get_context,
         cfd_->internal_stats()->GetFileReadHist(fp.GetHitFileLevel()),
@@ -1297,6 +1316,32 @@ void VersionStorageInfo::EstimateCompactionBytesNeeded(
   }
 }
 
+namespace {
+uint32_t GetExpiredTtlFilesCount(const ImmutableCFOptions& ioptions,
+                                 const std::vector<FileMetaData*>& files) {
+  uint32_t ttl_expired_files_count = 0;
+
+  int64_t _current_time;
+  auto status = ioptions.env->GetCurrentTime(&_current_time);
+  if (status.ok()) {
+    const uint64_t current_time = static_cast<uint64_t>(_current_time);
+    for (auto f : files) {
+      if (!f->being_compacted && f->fd.table_reader != nullptr &&
+          f->fd.table_reader->GetTableProperties() != nullptr) {
+        auto creation_time =
+            f->fd.table_reader->GetTableProperties()->creation_time;
+        if (creation_time > 0 &&
+            creation_time <
+                (current_time - ioptions.compaction_options_fifo.ttl)) {
+          ttl_expired_files_count++;
+        }
+      }
+    }
+  }
+  return ttl_expired_files_count;
+}
+}  // anonymous namespace
+
 void VersionStorageInfo::ComputeCompactionScore(
     const ImmutableCFOptions& immutable_cf_options,
     const MutableCFOptions& mutable_cf_options) {
@@ -1343,6 +1388,11 @@ void VersionStorageInfo::ComputeCompactionScore(
                   mutable_cf_options.level0_file_num_compaction_trigger,
               score);
         }
+        if (immutable_cf_options.compaction_options_fifo.ttl > 0) {
+          score = std::max(static_cast<double>(GetExpiredTtlFilesCount(
+                               immutable_cf_options, files_[level])),
+                           score);
+        }
 
       } else {
         score = static_cast<double>(num_sorted_runs) /
@@ -2199,13 +2249,16 @@ void Version::AddLiveFiles(std::vector<FileDescriptor>* live) {
   }
 }
 
-std::string Version::DebugString(bool hex) const {
+std::string Version::DebugString(bool hex, bool print_stats) const {
   std::string r;
   for (int level = 0; level < storage_info_.num_levels_; level++) {
     // E.g.,
     //   --- level 1 ---
     //   17:123['a' .. 'd']
     //   20:43['e' .. 'g']
+    //
+    // if print_stats=true:
+    //   17:123['a' .. 'd'](4096)
     r.append("--- level ");
     AppendNumberTo(&r, level);
     r.append(" --- version# ");
@@ -2221,7 +2274,14 @@ std::string Version::DebugString(bool hex) const {
       r.append(files[i]->smallest.DebugString(hex));
       r.append(" .. ");
       r.append(files[i]->largest.DebugString(hex));
-      r.append("]\n");
+      r.append("]");
+      if (print_stats) {
+        r.append("(");
+        r.append(ToString(
+            files[i]->stats.num_reads_sampled.load(std::memory_order_relaxed)));
+        r.append(")");
+      }
+      r.append("\n");
     }
   }
   return r;
@@ -2255,6 +2315,7 @@ VersionSet::VersionSet(const std::string& dbname,
       manifest_file_number_(0),  // Filled by Recover()
       pending_manifest_file_number_(0),
       last_sequence_(0),
+      last_to_be_written_sequence_(0),
       prev_log_number_(0),
       current_version_number_(0),
       manifest_file_size_(0),
@@ -2891,6 +2952,7 @@ Status VersionSet::Recover(
 
     manifest_file_size_ = current_manifest_file_size;
     next_file_number_.store(next_file + 1);
+    last_to_be_written_sequence_ = last_sequence;
     last_sequence_ = last_sequence;
     prev_log_number_ = previous_log_number;
 
@@ -3260,6 +3322,7 @@ Status VersionSet::DumpManifest(Options& options, std::string& dscname,
     }
 
     next_file_number_.store(next_file + 1);
+    last_to_be_written_sequence_ = last_sequence;
     last_sequence_ = last_sequence;
     prev_log_number_ = previous_log_number;
 
@@ -3544,7 +3607,8 @@ InternalIterator* VersionSet::MakeInputIterator(
                 false /* skip_filters */, (int)which /* level */,
                 range_del_agg),
             new LevelFileNumIterator(cfd->internal_comparator(),
-                                     c->input_levels(which)));
+                                     c->input_levels(which),
+                                     false /* don't sample compaction */));
       }
     }
   }

db/version_set.h

@@ -502,7 +502,7 @@ class Version {
   void AddLiveFiles(std::vector<FileDescriptor>* live);
 
   // Return a human readable string that describes this version's contents.
-  std::string DebugString(bool hex = false) const;
+  std::string DebugString(bool hex = false, bool print_stats = false) const;
 
   // Returns the version nuber of this version
   uint64_t GetVersionNumber() const { return version_number_; }
@@ -697,12 +697,31 @@ class VersionSet {
     return last_sequence_.load(std::memory_order_acquire);
   }
 
+  // Note: memory_order_acquire must be sufficient.
+  uint64_t LastToBeWrittenSequence() const {
+    return last_to_be_written_sequence_.load(std::memory_order_seq_cst);
+  }
+
   // Set the last sequence number to s.
   void SetLastSequence(uint64_t s) {
     assert(s >= last_sequence_);
+    // Last visible seqeunce must always be less than last written seq
+    assert(!db_options_->concurrent_prepare ||
+           s <= last_to_be_written_sequence_);
     last_sequence_.store(s, std::memory_order_release);
   }
 
+  // Note: memory_order_release must be sufficient
+  void SetLastToBeWrittenSequence(uint64_t s) {
+    assert(s >= last_to_be_written_sequence_);
+    last_to_be_written_sequence_.store(s, std::memory_order_seq_cst);
+  }
+
+  // Note: memory_order_release must be sufficient
+  uint64_t FetchAddLastToBeWrittenSequence(uint64_t s) {
+    return last_to_be_written_sequence_.fetch_add(s, std::memory_order_seq_cst);
+  }
+
   // Mark the specified file number as used.
   // REQUIRED: this is only called during single-threaded recovery
   void MarkFileNumberUsedDuringRecovery(uint64_t number);
@@ -802,7 +821,10 @@ class VersionSet {
   uint64_t manifest_file_number_;
   uint64_t options_file_number_;
   uint64_t pending_manifest_file_number_;
+  // The last seq visible to reads
   std::atomic<uint64_t> last_sequence_;
+  // The last seq with which a writer has written/will write.
+  std::atomic<uint64_t> last_to_be_written_sequence_;
   uint64_t prev_log_number_;  // 0 or backing store for memtable being compacted
 
   // Opened lazily

db/wal_manager_test.cc

@@ -67,6 +67,7 @@ class WalManagerTest : public testing::Test {
     batch.Put(key, value);
     WriteBatchInternal::SetSequence(&batch, seq);
     current_log_writer_->AddRecord(WriteBatchInternal::Contents(&batch));
+    versions_->SetLastToBeWrittenSequence(seq);
     versions_->SetLastSequence(seq);
   }
 

db/write_callback_test.cc

@@ -123,176 +123,180 @@ TEST_F(WriteCallbackTest, WriteWithCallbackTest) {
       {false, false, true, false, true},
   };
 
-  for (auto& allow_parallel : {true, false}) {
-    for (auto& allow_batching : {true, false}) {
-      for (auto& enable_WAL : {true, false}) {
-        for (auto& enable_pipelined_write : {true, false}) {
-          for (auto& write_group : write_scenarios) {
-            Options options;
-            options.create_if_missing = true;
-            options.allow_concurrent_memtable_write = allow_parallel;
-            options.enable_pipelined_write = enable_pipelined_write;
+  for (auto& two_queues : {true, false}) {
+    for (auto& allow_parallel : {true, false}) {
+      for (auto& allow_batching : {true, false}) {
+        for (auto& enable_WAL : {true, false}) {
+          for (auto& enable_pipelined_write : {true, false}) {
+            for (auto& write_group : write_scenarios) {
+              Options options;
+              options.create_if_missing = true;
+              options.allow_concurrent_memtable_write = allow_parallel;
+              options.enable_pipelined_write = enable_pipelined_write;
+              options.concurrent_prepare = two_queues;
 
-            ReadOptions read_options;
-            DB* db;
-            DBImpl* db_impl;
+              ReadOptions read_options;
+              DB* db;
+              DBImpl* db_impl;
 
-            DestroyDB(dbname, options);
-            ASSERT_OK(DB::Open(options, dbname, &db));
+              DestroyDB(dbname, options);
+              ASSERT_OK(DB::Open(options, dbname, &db));
 
-            db_impl = dynamic_cast<DBImpl*>(db);
-            ASSERT_TRUE(db_impl);
+              db_impl = dynamic_cast<DBImpl*>(db);
+              ASSERT_TRUE(db_impl);
 
-            std::atomic<uint64_t> threads_waiting(0);
-            std::atomic<uint64_t> seq(db_impl->GetLatestSequenceNumber());
-            ASSERT_EQ(db_impl->GetLatestSequenceNumber(), 0);
+              std::atomic<uint64_t> threads_waiting(0);
+              std::atomic<uint64_t> seq(db_impl->GetLatestSequenceNumber());
+              ASSERT_EQ(db_impl->GetLatestSequenceNumber(), 0);
 
-            rocksdb::SyncPoint::GetInstance()->SetCallBack(
-                "WriteThread::JoinBatchGroup:Wait", [&](void* arg) {
-                  uint64_t cur_threads_waiting = 0;
-                  bool is_leader = false;
-                  bool is_last = false;
+              rocksdb::SyncPoint::GetInstance()->SetCallBack(
+                  "WriteThread::JoinBatchGroup:Wait", [&](void* arg) {
+                    uint64_t cur_threads_waiting = 0;
+                    bool is_leader = false;
+                    bool is_last = false;
 
-                  // who am i
+                    // who am i
+                    do {
+                      cur_threads_waiting = threads_waiting.load();
+                      is_leader = (cur_threads_waiting == 0);
+                      is_last = (cur_threads_waiting == write_group.size() - 1);
+                    } while (!threads_waiting.compare_exchange_strong(
+                        cur_threads_waiting, cur_threads_waiting + 1));
+
+                    // check my state
+                    auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
+
+                    if (is_leader) {
+                      ASSERT_TRUE(writer->state ==
+                                  WriteThread::State::STATE_GROUP_LEADER);
+                    } else {
+                      ASSERT_TRUE(writer->state ==
+                                  WriteThread::State::STATE_INIT);
+                    }
+
+                    // (meta test) the first WriteOP should indeed be the first
+                    // and the last should be the last (all others can be out of
+                    // order)
+                    if (is_leader) {
+                      ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
+                                  !write_group.front().callback_.should_fail_);
+                    } else if (is_last) {
+                      ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
+                                  !write_group.back().callback_.should_fail_);
+                    }
+
+                    // wait for friends
+                    while (threads_waiting.load() < write_group.size()) {
+                    }
+                  });
+
+              rocksdb::SyncPoint::GetInstance()->SetCallBack(
+                  "WriteThread::JoinBatchGroup:DoneWaiting", [&](void* arg) {
+                    // check my state
+                    auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
+
+                    if (!allow_batching) {
+                      // no batching so everyone should be a leader
+                      ASSERT_TRUE(writer->state ==
+                                  WriteThread::State::STATE_GROUP_LEADER);
+                    } else if (!allow_parallel) {
+                      ASSERT_TRUE(writer->state ==
+                                      WriteThread::State::STATE_COMPLETED ||
+                                  (enable_pipelined_write &&
+                                   writer->state ==
+                                       WriteThread::State::
+                                           STATE_MEMTABLE_WRITER_LEADER));
+                    }
+                  });
+
+              std::atomic<uint32_t> thread_num(0);
+              std::atomic<char> dummy_key(0);
+              std::function<void()> write_with_callback_func = [&]() {
+                uint32_t i = thread_num.fetch_add(1);
+                Random rnd(i);
+
+                // leaders gotta lead
+                while (i > 0 && threads_waiting.load() < 1) {
+                }
+
+                // loser has to lose
+                while (i == write_group.size() - 1 &&
+                       threads_waiting.load() < write_group.size() - 1) {
+                }
+
+                auto& write_op = write_group.at(i);
+                write_op.Clear();
+                write_op.callback_.allow_batching_ = allow_batching;
+
+                // insert some keys
+                for (uint32_t j = 0; j < rnd.Next() % 50; j++) {
+                  // grab unique key
+                  char my_key = 0;
                   do {
-                    cur_threads_waiting = threads_waiting.load();
-                    is_leader = (cur_threads_waiting == 0);
-                    is_last = (cur_threads_waiting == write_group.size() - 1);
-                  } while (!threads_waiting.compare_exchange_strong(
-                      cur_threads_waiting, cur_threads_waiting + 1));
+                    my_key = dummy_key.load();
+                  } while (
+                      !dummy_key.compare_exchange_strong(my_key, my_key + 1));
 
-                  // check my state
-                  auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
+                  string skey(5, my_key);
+                  string sval(10, my_key);
+                  write_op.Put(skey, sval);
 
-                  if (is_leader) {
-                    ASSERT_TRUE(writer->state ==
-                                WriteThread::State::STATE_GROUP_LEADER);
-                  } else {
-                    ASSERT_TRUE(writer->state ==
-                                WriteThread::State::STATE_INIT);
+                  if (!write_op.callback_.should_fail_) {
+                    seq.fetch_add(1);
                   }
-
-                  // (meta test) the first WriteOP should indeed be the first
-                  // and the last should be the last (all others can be out of
-                  // order)
-                  if (is_leader) {
-                    ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
-                                !write_group.front().callback_.should_fail_);
-                  } else if (is_last) {
-                    ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
-                                !write_group.back().callback_.should_fail_);
-                  }
-
-                  // wait for friends
-                  while (threads_waiting.load() < write_group.size()) {
-                  }
-                });
-
-            rocksdb::SyncPoint::GetInstance()->SetCallBack(
-                "WriteThread::JoinBatchGroup:DoneWaiting", [&](void* arg) {
-                  // check my state
-                  auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
-
-                  if (!allow_batching) {
-                    // no batching so everyone should be a leader
-                    ASSERT_TRUE(writer->state ==
-                                WriteThread::State::STATE_GROUP_LEADER);
-                  } else if (!allow_parallel) {
-                    ASSERT_TRUE(
-                        writer->state == WriteThread::State::STATE_COMPLETED ||
-                        (enable_pipelined_write &&
-                         writer->state ==
-                             WriteThread::State::STATE_MEMTABLE_WRITER_LEADER));
-                  }
-                });
-
-            std::atomic<uint32_t> thread_num(0);
-            std::atomic<char> dummy_key(0);
-            std::function<void()> write_with_callback_func = [&]() {
-              uint32_t i = thread_num.fetch_add(1);
-              Random rnd(i);
-
-              // leaders gotta lead
-              while (i > 0 && threads_waiting.load() < 1) {
-              }
-
-              // loser has to lose
-              while (i == write_group.size() - 1 &&
-                     threads_waiting.load() < write_group.size() - 1) {
-              }
-
-              auto& write_op = write_group.at(i);
-              write_op.Clear();
-              write_op.callback_.allow_batching_ = allow_batching;
-
-              // insert some keys
-              for (uint32_t j = 0; j < rnd.Next() % 50; j++) {
-                // grab unique key
-                char my_key = 0;
-                do {
-                  my_key = dummy_key.load();
-                } while (
-                    !dummy_key.compare_exchange_strong(my_key, my_key + 1));
-
-                string skey(5, my_key);
-                string sval(10, my_key);
-                write_op.Put(skey, sval);
-
-                if (!write_op.callback_.should_fail_) {
-                  seq.fetch_add(1);
                 }
-              }
 
-              WriteOptions woptions;
-              woptions.disableWAL = !enable_WAL;
-              woptions.sync = enable_WAL;
-              Status s = db_impl->WriteWithCallback(
-                  woptions, &write_op.write_batch_, &write_op.callback_);
+                WriteOptions woptions;
+                woptions.disableWAL = !enable_WAL;
+                woptions.sync = enable_WAL;
+                Status s = db_impl->WriteWithCallback(
+                    woptions, &write_op.write_batch_, &write_op.callback_);
 
-              if (write_op.callback_.should_fail_) {
-                ASSERT_TRUE(s.IsBusy());
-              } else {
-                ASSERT_OK(s);
-              }
-            };
-
-            rocksdb::SyncPoint::GetInstance()->EnableProcessing();
-
-            // do all the writes
-            std::vector<port::Thread> threads;
-            for (uint32_t i = 0; i < write_group.size(); i++) {
-              threads.emplace_back(write_with_callback_func);
-            }
-            for (auto& t : threads) {
-              t.join();
-            }
-
-            rocksdb::SyncPoint::GetInstance()->DisableProcessing();
-
-            // check for keys
-            string value;
-            for (auto& w : write_group) {
-              ASSERT_TRUE(w.callback_.was_called_.load());
-              for (auto& kvp : w.kvs_) {
-                if (w.callback_.should_fail_) {
-                  ASSERT_TRUE(
-                      db->Get(read_options, kvp.first, &value).IsNotFound());
+                if (write_op.callback_.should_fail_) {
+                  ASSERT_TRUE(s.IsBusy());
                 } else {
-                  ASSERT_OK(db->Get(read_options, kvp.first, &value));
-                  ASSERT_EQ(value, kvp.second);
+                  ASSERT_OK(s);
+                }
+              };
+
+              rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
+              // do all the writes
+              std::vector<port::Thread> threads;
+              for (uint32_t i = 0; i < write_group.size(); i++) {
+                threads.emplace_back(write_with_callback_func);
+              }
+              for (auto& t : threads) {
+                t.join();
+              }
+
+              rocksdb::SyncPoint::GetInstance()->DisableProcessing();
+
+              // check for keys
+              string value;
+              for (auto& w : write_group) {
+                ASSERT_TRUE(w.callback_.was_called_.load());
+                for (auto& kvp : w.kvs_) {
+                  if (w.callback_.should_fail_) {
+                    ASSERT_TRUE(
+                        db->Get(read_options, kvp.first, &value).IsNotFound());
+                  } else {
+                    ASSERT_OK(db->Get(read_options, kvp.first, &value));
+                    ASSERT_EQ(value, kvp.second);
+                  }
                 }
               }
+
+              ASSERT_EQ(seq.load(), db_impl->GetLatestSequenceNumber());
+
+              delete db;
+              DestroyDB(dbname, options);
             }
-
-            ASSERT_EQ(seq.load(), db_impl->GetLatestSequenceNumber());
-
-            delete db;
-            DestroyDB(dbname, options);
           }
         }
       }
     }
-  }
+}
 }
 
 TEST_F(WriteCallbackTest, WriteCallBackTest) {

docs/_data/authors.yml

@@ -56,3 +56,7 @@ ajkr:
 sagar0:
   full_name: Sagar Vemuri
   fbid: 2419111
+
+lightmark:
+  full_name: Aaron Gao
+  fbid: 1351549072

docs/_posts/2017-06-26-17-level-based-changes.markdown

@@ -0,0 +1,60 @@
+---
+title: Level-based Compaction Changes
+layout: post
+author: ajkr
+category: blog
+---
+
+### Introduction
+
+RocksDB provides an option to limit the number of L0 files, which bounds read-amplification. Since L0 files (unlike files at lower levels) can span the entire key-range, a key might be in any file, thus reads need to check them one-by-one. Users often wish to configure a low limit to improve their read latency.
+
+Although, the mechanism with which we enforce L0's file count limit may be unappealing. When the limit is reached, RocksDB intentionally delays user writes. This slows down accumulation of files in L0, and frees up resources for compacting files down to lower levels. But adding delays will significantly increase user-visible write latency jitter.
+
+Also, due to how L0 files can span the entire key-range, compaction parallelization is limited. Files at L0 or L1 may be locked due to involvement in pending L0->L1 or L1->L2 compactions. We can only schedule a parallel L0->L1 compaction if it does not require any of the locked files, which is typically not the case.
+
+To handle these constraints better, we added a new type of compaction, L0->L0. It quickly reduces file count in L0 and can be scheduled even when L1 files are locked, unlike L0->L1. We also changed the L0->L1 picking algorithm to increase opportunities for parallelism.
+
+### Old L0->L1 Picking Logic
+
+Previously, our logic for picking which L0 file to compact was the same as every other level: pick the largest file in the level. One special property of L0->L1 compaction is that files can overlap in the input level, so those overlapping files must be pulled in as well. For example, a compaction may look like this:
+
+![full-range.png](/static/images/compaction/full-range.png)
+
+This compaction pulls in every L0 and L1 file. This happens regardless of which L0 file is initially chosen as each file overlaps with every other file.
+
+Users may insert their data less uniformly in the key-range. For example, a database may look like this during L0->L1 compaction:
+
+![part-range-old.png](/static/images/compaction/part-range-old.png)
+
+Let's say the third file from the top is the largest, and let's say the top two files are created after the compaction started. When the compaction is picked, the fourth L0 file and six rightmost L1 files are pulled in due to overlap. Notice this leaves the database in a state where we might not be able to schedule parallel compactions. For example, if the sixth file from the top is the next largest, we can't compact it because it overlaps with the top two files, which overlap with the locked L0 files.
+
+We can now see the high-level problems with this approach more clearly. First, locked files in L0 or L1 prevent us from parallelizing compactions. When locked files block L0->L1 compaction, there is nothing we can do to eliminate L0 files. Second, L0->L1 compactions are relatively slow. As we saw, when keys are uniformly distributed, L0->L1 compacts two entire levels. While this is happening, new files are being flushed to L0, advancing towards the file count limit.
+
+### New L0->L0 Algorithm
+
+We introduced compaction within L0 to improve both parallelization and speed of reducing L0 file count. An L0->L0 compaction may look like this:
+
+![l1-l2-contend.png](/static/images/compaction/l1-l2-contend.png)
+
+Say the L1->L2 compaction started first. Now L0->L1 is prevented by the locked L1 file. In this case, we compact files within L0. This allows us to start the work for eliminating L0 files earlier. It also lets us do less work since we don't pull in any L1 files, whereas L0->L1 compaction would've pulled in all of them. This lets us quickly reduce L0 file count to keep read-amp low while sustaining large bursts of writes (i.e., fast accumulation of L0 files).
+
+The tradeoff is this increases total compaction work, as we're now compacting files without contributing towards our eventual goal of moving them towards lower levels. Our benchmarks, though, consistently show less compaction stalls and improved write throughput. One justification is that L0 file data is highly likely in page cache and/or block cache due to it being recently written and frequently accessed. So, this type of compaction is relatively cheap compared to compactions at lower levels.
+
+This feature is available since RocksDB 5.4.
+
+### New L0->L1 Picking Logic
+
+Recall how the old L0->L1 picking algorithm chose the largest L0 file for compaction. This didn't fit well with L0->L0 compaction, which operates on a span of files. That span begins at the newest L0 file, and expands towards older files as long as they're not being compacted. Since the largest file may be anywhere, the old L0->L1 picking logic could arbitrarily prevent us from getting a long span of files. See the second illustration in this post for a scenario where this would happen.
+
+So, we changed the L0->L1 picking algorithm to start from the oldest file and expand towards newer files as long as they're not being compacted. For example:
+
+![l0-l1-contend.png](/static/images/compaction/l0-l1-contend.png)
+
+Now, there can never be L0 files unreachable for L0->L0 due to L0->L1 selecting files in the middle. When longer spans of files are available for L0->L0, we perform less compaction work per deleted L0 file, thus improving efficiency.
+
+This feature will be available in RocksDB 5.7.
+
+### Performance Changes
+
+Mark Callaghan did the most extensive benchmarking of this feature's impact on MyRocks. See his results [here](http://smalldatum.blogspot.com/2017/05/innodb-myrocks-and-tokudb-on-insert.html). Note the primary change between his March 17 and April 14 builds is the latter performs L0->L0 compaction.

docs/_posts/2017-06-29-rocksdb-5-5-1-released.markdown

@@ -0,0 +1,22 @@
+---
+title: RocksDB 5.5.1 Released!
+layout: post
+author: lightmark
+category: blog
+---
+
+### New Features
+* FIFO compaction to support Intra L0 compaction too with CompactionOptionsFIFO.allow_compaction=true.
+* Statistics::Reset() to reset user stats.
+* ldb add option --try_load_options, which will open DB with its own option file.
+* Introduce WriteBatch::PopSavePoint to pop the most recent save point explicitly.
+* Support dynamically change `max_open_files` option via SetDBOptions()
+* Added DB::CreateColumnFamilie() and DB::DropColumnFamilies() to bulk create/drop column families.
+* Add debugging function `GetAllKeyVersions` to see internal versions of a range of keys.
+* Support file ingestion with universal compaction style
+* Support file ingestion behind with option `allow_ingest_behind`
+* New option enable_pipelined_write which may improve write throughput in case writing from multiple threads and WAL enabled.
+
+### Bug Fixes
+* Fix the bug that Direct I/O uses direct reads for non-SST file
+* Fix the bug that flush doesn't respond to fsync result

env/env_encryption.cc

@@ -0,0 +1,911 @@
+//  Copyright (c) 2011-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//  This source code is also licensed under the GPLv2 license found in the
+//  COPYING file in the root directory of this source tree.
+
+#ifndef ROCKSDB_LITE
+
+#include <algorithm>
+#include <cctype>
+#include <iostream>
+
+#include "rocksdb/env_encryption.h"
+#include "util/aligned_buffer.h"
+#include "util/coding.h"
+#include "util/random.h"
+
+#endif
+
+namespace rocksdb {
+
+#ifndef ROCKSDB_LITE
+
+class EncryptedSequentialFile : public SequentialFile {
+  private:
+    std::unique_ptr<SequentialFile> file_;
+    std::unique_ptr<BlockAccessCipherStream> stream_;
+    uint64_t offset_;
+    size_t prefixLength_;
+
+     public:
+  // Default ctor. Given underlying sequential file is supposed to be at
+  // offset == prefixLength.
+  EncryptedSequentialFile(SequentialFile* f, BlockAccessCipherStream* s, size_t prefixLength)
+      : file_(f), stream_(s), offset_(prefixLength), prefixLength_(prefixLength) {
+  }
+
+  // Read up to "n" bytes from the file.  "scratch[0..n-1]" may be
+  // written by this routine.  Sets "*result" to the data that was
+  // read (including if fewer than "n" bytes were successfully read).
+  // May set "*result" to point at data in "scratch[0..n-1]", so
+  // "scratch[0..n-1]" must be live when "*result" is used.
+  // If an error was encountered, returns a non-OK status.
+  //
+  // REQUIRES: External synchronization
+  virtual Status Read(size_t n, Slice* result, char* scratch) override {
+    assert(scratch);
+    Status status = file_->Read(n, result, scratch);
+    if (!status.ok()) {
+      return status;
+    }
+    status = stream_->Decrypt(offset_, (char*)result->data(), result->size());
+    offset_ += result->size(); // We've already ready data from disk, so update offset_ even if decryption fails.
+    return status;
+  }
+
+  // Skip "n" bytes from the file. This is guaranteed to be no
+  // slower that reading the same data, but may be faster.
+  //
+  // If end of file is reached, skipping will stop at the end of the
+  // file, and Skip will return OK.
+  //
+  // REQUIRES: External synchronization
+  virtual Status Skip(uint64_t n) override {
+    auto status = file_->Skip(n);
+    if (!status.ok()) {
+      return status;
+    }
+    offset_ += n;
+    return status;
+  }
+
+  // Indicates the upper layers if the current SequentialFile implementation
+  // uses direct IO.
+  virtual bool use_direct_io() const override { 
+    return file_->use_direct_io(); 
+  }
+
+  // Use the returned alignment value to allocate
+  // aligned buffer for Direct I/O
+  virtual size_t GetRequiredBufferAlignment() const override { 
+    return file_->GetRequiredBufferAlignment(); 
+  }
+
+  // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  virtual Status InvalidateCache(size_t offset, size_t length) override {
+    return file_->InvalidateCache(offset + prefixLength_, length);
+  }
+
+  // Positioned Read for direct I/O
+  // If Direct I/O enabled, offset, n, and scratch should be properly aligned
+  virtual Status PositionedRead(uint64_t offset, size_t n, Slice* result, char* scratch) override {
+    assert(scratch);
+    offset += prefixLength_; // Skip prefix
+    auto status = file_->PositionedRead(offset, n, result, scratch);
+    if (!status.ok()) {
+      return status;
+    }
+    offset_ = offset + result->size();
+    status = stream_->Decrypt(offset, (char*)result->data(), result->size());
+    return status;
+  }
+
+};
+
+// A file abstraction for randomly reading the contents of a file.
+class EncryptedRandomAccessFile : public RandomAccessFile {
+  private:
+    std::unique_ptr<RandomAccessFile> file_;
+    std::unique_ptr<BlockAccessCipherStream> stream_;
+    size_t prefixLength_;
+
+ public:
+  EncryptedRandomAccessFile(RandomAccessFile* f, BlockAccessCipherStream* s, size_t prefixLength)
+    : file_(f), stream_(s), prefixLength_(prefixLength) { }
+
+  // Read up to "n" bytes from the file starting at "offset".
+  // "scratch[0..n-1]" may be written by this routine.  Sets "*result"
+  // to the data that was read (including if fewer than "n" bytes were
+  // successfully read).  May set "*result" to point at data in
+  // "scratch[0..n-1]", so "scratch[0..n-1]" must be live when
+  // "*result" is used.  If an error was encountered, returns a non-OK
+  // status.
+  //
+  // Safe for concurrent use by multiple threads.
+  // If Direct I/O enabled, offset, n, and scratch should be aligned properly.
+  virtual Status Read(uint64_t offset, size_t n, Slice* result, char* scratch) const override {
+    assert(scratch);
+    offset += prefixLength_;
+    auto status = file_->Read(offset, n, result, scratch);
+    if (!status.ok()) {
+      return status;
+    }
+    status = stream_->Decrypt(offset, (char*)result->data(), result->size());
+    return status;
+  }
+
+  // Readahead the file starting from offset by n bytes for caching.
+  virtual Status Prefetch(uint64_t offset, size_t n) override {
+    //return Status::OK();
+    return file_->Prefetch(offset + prefixLength_, n);
+  }
+
+  // Tries to get an unique ID for this file that will be the same each time
+  // the file is opened (and will stay the same while the file is open).
+  // Furthermore, it tries to make this ID at most "max_size" bytes. If such an
+  // ID can be created this function returns the length of the ID and places it
+  // in "id"; otherwise, this function returns 0, in which case "id"
+  // may not have been modified.
+  //
+  // This function guarantees, for IDs from a given environment, two unique ids
+  // cannot be made equal to eachother by adding arbitrary bytes to one of
+  // them. That is, no unique ID is the prefix of another.
+  //
+  // This function guarantees that the returned ID will not be interpretable as
+  // a single varint.
+  //
+  // Note: these IDs are only valid for the duration of the process.
+  virtual size_t GetUniqueId(char* id, size_t max_size) const override {
+    return file_->GetUniqueId(id, max_size);
+  };
+
+  virtual void Hint(AccessPattern pattern) override {
+    file_->Hint(pattern);
+  }
+
+  // Indicates the upper layers if the current RandomAccessFile implementation
+  // uses direct IO.
+  virtual bool use_direct_io() const override {
+     return file_->use_direct_io(); 
+  }
+
+  // Use the returned alignment value to allocate
+  // aligned buffer for Direct I/O
+  virtual size_t GetRequiredBufferAlignment() const override { 
+    return file_->GetRequiredBufferAlignment(); 
+  }
+
+  // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  virtual Status InvalidateCache(size_t offset, size_t length) override {
+    return file_->InvalidateCache(offset + prefixLength_, length);
+  }
+};
+
+// A file abstraction for sequential writing.  The implementation
+// must provide buffering since callers may append small fragments
+// at a time to the file.
+class EncryptedWritableFile : public WritableFileWrapper {
+  private:
+    std::unique_ptr<WritableFile> file_;
+    std::unique_ptr<BlockAccessCipherStream> stream_;
+    size_t prefixLength_;
+
+ public:
+  // Default ctor. Prefix is assumed to be written already.
+  EncryptedWritableFile(WritableFile* f, BlockAccessCipherStream* s, size_t prefixLength)
+    : WritableFileWrapper(f), file_(f), stream_(s), prefixLength_(prefixLength) { }
+
+  Status Append(const Slice& data) override { 
+    AlignedBuffer buf;
+    Status status;
+    Slice dataToAppend(data); 
+    if (data.size() > 0) {
+      auto offset = file_->GetFileSize(); // size including prefix
+      // Encrypt in cloned buffer
+      buf.Alignment(GetRequiredBufferAlignment());
+      buf.AllocateNewBuffer(data.size());
+      memmove(buf.BufferStart(), data.data(), data.size());
+      status = stream_->Encrypt(offset, buf.BufferStart(), data.size());
+      if (!status.ok()) {
+        return status;
+      }
+      dataToAppend = Slice(buf.BufferStart(), data.size());
+    }
+    status = file_->Append(dataToAppend); 
+    if (!status.ok()) {
+      return status;
+    }
+    return status;
+  }
+
+  Status PositionedAppend(const Slice& data, uint64_t offset) override {
+    AlignedBuffer buf;
+    Status status;
+    Slice dataToAppend(data); 
+    offset += prefixLength_;
+    if (data.size() > 0) {
+      // Encrypt in cloned buffer
+      buf.Alignment(GetRequiredBufferAlignment());
+      buf.AllocateNewBuffer(data.size());
+      memmove(buf.BufferStart(), data.data(), data.size());
+      status = stream_->Encrypt(offset, buf.BufferStart(), data.size());
+      if (!status.ok()) {
+        return status;
+      }
+      dataToAppend = Slice(buf.BufferStart(), data.size());
+    }
+    status = file_->PositionedAppend(dataToAppend, offset);
+    if (!status.ok()) {
+      return status;
+    }
+    return status;
+  }
+
+  // Indicates the upper layers if the current WritableFile implementation
+  // uses direct IO.
+  virtual bool use_direct_io() const override { return file_->use_direct_io(); }
+
+  // Use the returned alignment value to allocate
+  // aligned buffer for Direct I/O
+  virtual size_t GetRequiredBufferAlignment() const override { return file_->GetRequiredBufferAlignment(); } 
+
+    /*
+   * Get the size of valid data in the file.
+   */
+  virtual uint64_t GetFileSize() override {
+    return file_->GetFileSize() - prefixLength_;
+  }
+
+  // Truncate is necessary to trim the file to the correct size
+  // before closing. It is not always possible to keep track of the file
+  // size due to whole pages writes. The behavior is undefined if called
+  // with other writes to follow.
+  virtual Status Truncate(uint64_t size) override {
+    return file_->Truncate(size + prefixLength_);
+  }
+
+    // Remove any kind of caching of data from the offset to offset+length
+  // of this file. If the length is 0, then it refers to the end of file.
+  // If the system is not caching the file contents, then this is a noop.
+  // This call has no effect on dirty pages in the cache.
+  virtual Status InvalidateCache(size_t offset, size_t length) override {
+    return file_->InvalidateCache(offset + prefixLength_, length);
+  }
+
+  // Sync a file range with disk.
+  // offset is the starting byte of the file range to be synchronized.
+  // nbytes specifies the length of the range to be synchronized.
+  // This asks the OS to initiate flushing the cached data to disk,
+  // without waiting for completion.
+  // Default implementation does nothing.
+  virtual Status RangeSync(uint64_t offset, uint64_t nbytes) override { 
+    return file_->RangeSync(offset + prefixLength_, nbytes);
+  }
+
+  // PrepareWrite performs any necessary preparation for a write
+  // before the write actually occurs.  This allows for pre-allocation
+  // of space on devices where it can result in less file
+  // fragmentation and/or less waste from over-zealous filesystem
+  // pre-allocation.
+  virtual void PrepareWrite(size_t offset, size_t len) override {
+    file_->PrepareWrite(offset + prefixLength_, len);
+  }
+
+  // Pre-allocates space for a file.
+  virtual Status Allocate(uint64_t offset, uint64_t len) override {
+    return file_->Allocate(offset + prefixLength_, len);
+  }
+};
+
+// A file abstraction for random reading and writing.
+class EncryptedRandomRWFile : public RandomRWFile {
+  private:
+    std::unique_ptr<RandomRWFile> file_;
+    std::unique_ptr<BlockAccessCipherStream> stream_;
+    size_t prefixLength_;
+
+ public:
+  EncryptedRandomRWFile(RandomRWFile* f, BlockAccessCipherStream* s, size_t prefixLength)
+    : file_(f), stream_(s), prefixLength_(prefixLength) {}
+
+  // Indicates if the class makes use of direct I/O
+  // If false you must pass aligned buffer to Write()
+  virtual bool use_direct_io() const override { return file_->use_direct_io(); }
+
+  // Use the returned alignment value to allocate
+  // aligned buffer for Direct I/O
+  virtual size_t GetRequiredBufferAlignment() const override { 
+    return file_->GetRequiredBufferAlignment(); 
+  }
+
+  // Write bytes in `data` at  offset `offset`, Returns Status::OK() on success.
+  // Pass aligned buffer when use_direct_io() returns true.
+  virtual Status Write(uint64_t offset, const Slice& data) override {
+    AlignedBuffer buf;
+    Status status;
+    Slice dataToWrite(data); 
+    offset += prefixLength_;
+    if (data.size() > 0) {
+      // Encrypt in cloned buffer
+      buf.Alignment(GetRequiredBufferAlignment());
+      buf.AllocateNewBuffer(data.size());
+      memmove(buf.BufferStart(), data.data(), data.size());
+      status = stream_->Encrypt(offset, buf.BufferStart(), data.size());
+      if (!status.ok()) {
+        return status;
+      }
+      dataToWrite = Slice(buf.BufferStart(), data.size());
+    }
+    status = file_->Write(offset, dataToWrite);
+    return status;
+  }
+
+  // Read up to `n` bytes starting from offset `offset` and store them in
+  // result, provided `scratch` size should be at least `n`.
+  // Returns Status::OK() on success.
+  virtual Status Read(uint64_t offset, size_t n, Slice* result, char* scratch) const override { 
+    assert(scratch);
+    offset += prefixLength_;
+    auto status = file_->Read(offset, n, result, scratch);
+    if (!status.ok()) {
+      return status;
+    }
+    status = stream_->Decrypt(offset, (char*)result->data(), result->size());
+    return status;
+  }
+
+  virtual Status Flush() override {
+    return file_->Flush();
+  }
+
+  virtual Status Sync() override {
+    return file_->Sync();
+  }
+
+  virtual Status Fsync() override { 
+    return file_->Fsync();
+  }
+
+  virtual Status Close() override {
+    return file_->Close();
+  }
+};
+
+// EncryptedEnv implements an Env wrapper that adds encryption to files stored on disk.
+class EncryptedEnv : public EnvWrapper {
+ public:
+  EncryptedEnv(Env* base_env, EncryptionProvider *provider)
+      : EnvWrapper(base_env) {
+    provider_ = provider;
+  }
+
+  // NewSequentialFile opens a file for sequential reading.
+  virtual Status NewSequentialFile(const std::string& fname,
+                                   std::unique_ptr<SequentialFile>* result,
+                                   const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_reads) {
+      return Status::InvalidArgument();
+    }
+    // Open file using underlying Env implementation
+    std::unique_ptr<SequentialFile> underlying;
+    auto status = EnvWrapper::NewSequentialFile(fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Read prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      // Read prefix 
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      status = underlying->Read(prefixLength, &prefixSlice, prefixBuf.BufferStart());
+      if (!status.ok()) {
+        return status;
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<SequentialFile>(new EncryptedSequentialFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+
+  // NewRandomAccessFile opens a file for random read access.
+  virtual Status NewRandomAccessFile(const std::string& fname,
+                                     unique_ptr<RandomAccessFile>* result,
+                                     const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_reads) {
+      return Status::InvalidArgument();
+    }
+    // Open file using underlying Env implementation
+    std::unique_ptr<RandomAccessFile> underlying;
+    auto status = EnvWrapper::NewRandomAccessFile(fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Read prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      // Read prefix 
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      status = underlying->Read(0, prefixLength, &prefixSlice, prefixBuf.BufferStart());
+      if (!status.ok()) {
+        return status;
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<RandomAccessFile>(new EncryptedRandomAccessFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+  
+  // NewWritableFile opens a file for sequential writing.
+  virtual Status NewWritableFile(const std::string& fname,
+                                 unique_ptr<WritableFile>* result,
+                                 const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_writes) {
+      return Status::InvalidArgument();
+    }
+    // Open file using underlying Env implementation
+    std::unique_ptr<WritableFile> underlying;
+    Status status = EnvWrapper::NewWritableFile(fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Initialize & write prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      // Initialize prefix 
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      provider_->CreateNewPrefix(fname, prefixBuf.BufferStart(), prefixLength);
+      prefixSlice = Slice(prefixBuf.BufferStart(), prefixLength);
+      // Write prefix 
+      status = underlying->Append(prefixSlice);
+      if (!status.ok()) {
+        return status;
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<WritableFile>(new EncryptedWritableFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+
+  // Create an object that writes to a new file with the specified
+  // name.  Deletes any existing file with the same name and creates a
+  // new file.  On success, stores a pointer to the new file in
+  // *result and returns OK.  On failure stores nullptr in *result and
+  // returns non-OK.
+  //
+  // The returned file will only be accessed by one thread at a time.
+  virtual Status ReopenWritableFile(const std::string& fname,
+                                   unique_ptr<WritableFile>* result,
+                                   const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_writes) {
+      return Status::InvalidArgument();
+    }
+    // Open file using underlying Env implementation
+    std::unique_ptr<WritableFile> underlying;
+    Status status = EnvWrapper::ReopenWritableFile(fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Initialize & write prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      // Initialize prefix 
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      provider_->CreateNewPrefix(fname, prefixBuf.BufferStart(), prefixLength);
+      prefixSlice = Slice(prefixBuf.BufferStart(), prefixLength);
+      // Write prefix 
+      status = underlying->Append(prefixSlice);
+      if (!status.ok()) {
+        return status;
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<WritableFile>(new EncryptedWritableFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+
+  // Reuse an existing file by renaming it and opening it as writable.
+  virtual Status ReuseWritableFile(const std::string& fname,
+                                   const std::string& old_fname,
+                                   unique_ptr<WritableFile>* result,
+                                   const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_writes) {
+      return Status::InvalidArgument();
+    }
+    // Open file using underlying Env implementation
+    std::unique_ptr<WritableFile> underlying;
+    Status status = EnvWrapper::ReuseWritableFile(fname, old_fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Initialize & write prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      // Initialize prefix 
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      provider_->CreateNewPrefix(fname, prefixBuf.BufferStart(), prefixLength);
+      prefixSlice = Slice(prefixBuf.BufferStart(), prefixLength);
+      // Write prefix 
+      status = underlying->Append(prefixSlice);
+      if (!status.ok()) {
+        return status;
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<WritableFile>(new EncryptedWritableFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+
+  // Open `fname` for random read and write, if file dont exist the file
+  // will be created.  On success, stores a pointer to the new file in
+  // *result and returns OK.  On failure returns non-OK.
+  //
+  // The returned file will only be accessed by one thread at a time.
+  virtual Status NewRandomRWFile(const std::string& fname,
+                                 unique_ptr<RandomRWFile>* result,
+                                 const EnvOptions& options) override {
+    result->reset();
+    if (options.use_mmap_reads || options.use_mmap_writes) {
+      return Status::InvalidArgument();
+    }
+    // Check file exists
+    bool isNewFile = !FileExists(fname).ok();
+
+    // Open file using underlying Env implementation
+    std::unique_ptr<RandomRWFile> underlying;
+    Status status = EnvWrapper::NewRandomRWFile(fname, &underlying, options);
+    if (!status.ok()) {
+      return status;
+    }
+    // Read or Initialize & write prefix (if needed)
+    AlignedBuffer prefixBuf;
+    Slice prefixSlice;
+    size_t prefixLength = provider_->GetPrefixLength();
+    if (prefixLength > 0) {
+      prefixBuf.Alignment(underlying->GetRequiredBufferAlignment());
+      prefixBuf.AllocateNewBuffer(prefixLength);
+      if (!isNewFile) {
+        // File already exists, read prefix
+        status = underlying->Read(0, prefixLength, &prefixSlice, prefixBuf.BufferStart());
+        if (!status.ok()) {
+          return status;
+        }
+      } else {
+        // File is new, initialize & write prefix 
+        provider_->CreateNewPrefix(fname, prefixBuf.BufferStart(), prefixLength);
+        prefixSlice = Slice(prefixBuf.BufferStart(), prefixLength);
+        // Write prefix 
+        status = underlying->Write(0, prefixSlice);
+        if (!status.ok()) {
+          return status;
+        }
+      }
+    }
+    // Create cipher stream
+    std::unique_ptr<BlockAccessCipherStream> stream;
+    status = provider_->CreateCipherStream(fname, options, prefixSlice, &stream);
+    if (!status.ok()) {
+      return status;
+    }
+    (*result) = std::unique_ptr<RandomRWFile>(new EncryptedRandomRWFile(underlying.release(), stream.release(), prefixLength));
+    return Status::OK();
+  }
+
+    // Store in *result the attributes of the children of the specified directory.
+  // In case the implementation lists the directory prior to iterating the files
+  // and files are concurrently deleted, the deleted files will be omitted from
+  // result.
+  // The name attributes are relative to "dir".
+  // Original contents of *results are dropped.
+  // Returns OK if "dir" exists and "*result" contains its children.
+  //         NotFound if "dir" does not exist, the calling process does not have
+  //                  permission to access "dir", or if "dir" is invalid.
+  //         IOError if an IO Error was encountered
+  virtual Status GetChildrenFileAttributes(const std::string& dir, std::vector<FileAttributes>* result) override {
+    auto status = EnvWrapper::GetChildrenFileAttributes(dir, result);
+    if (!status.ok()) {
+      return status;
+    }
+    size_t prefixLength = provider_->GetPrefixLength();
+    for (auto it = std::begin(*result); it!=std::end(*result); ++it) {
+      assert(it->size_bytes >= prefixLength);
+      it->size_bytes -= prefixLength;
+    }
+    return Status::OK();
+ }
+
+  // Store the size of fname in *file_size.
+  virtual Status GetFileSize(const std::string& fname, uint64_t* file_size) override {
+    auto status = EnvWrapper::GetFileSize(fname, file_size);
+    if (!status.ok()) {
+      return status;
+    }
+    size_t prefixLength = provider_->GetPrefixLength();
+    assert(*file_size >= prefixLength);
+    *file_size -= prefixLength;
+    return Status::OK();    
+  }
+
+ private:
+  EncryptionProvider *provider_;
+};
+
+
+// Returns an Env that encrypts data when stored on disk and decrypts data when 
+// read from disk.
+Env* NewEncryptedEnv(Env* base_env, EncryptionProvider* provider) {
+  return new EncryptedEnv(base_env, provider);
+}
+
+// Encrypt one or more (partial) blocks of data at the file offset.
+// Length of data is given in dataSize.
+Status BlockAccessCipherStream::Encrypt(uint64_t fileOffset, char *data, size_t dataSize) {
+  // Calculate block index
+  auto blockSize = BlockSize();
+  uint64_t blockIndex = fileOffset / blockSize;
+  size_t blockOffset = fileOffset % blockSize;
+  unique_ptr<char[]> blockBuffer;
+
+  std::string scratch;
+  AllocateScratch(scratch);
+
+  // Encrypt individual blocks.
+  while (1) {
+    char *block = data;
+    size_t n = std::min(dataSize, blockSize - blockOffset);
+    if (n != blockSize) {
+      // We're not encrypting a full block. 
+      // Copy data to blockBuffer
+      if (!blockBuffer.get()) {
+        // Allocate buffer 
+        blockBuffer = unique_ptr<char[]>(new char[blockSize]);
+      }
+      block = blockBuffer.get();
+      // Copy plain data to block buffer 
+      memmove(block + blockOffset, data, n);
+    }
+    auto status = EncryptBlock(blockIndex, block, (char*)scratch.data());
+    if (!status.ok()) {
+      return status;
+    }
+    if (block != data) {
+      // Copy encrypted data back to `data`.
+      memmove(data, block + blockOffset, n);
+    }
+    dataSize -= n;
+    if (dataSize == 0) {
+      return Status::OK();
+    }
+    data += n;
+    blockOffset = 0;
+    blockIndex++;
+  }
+}
+
+// Decrypt one or more (partial) blocks of data at the file offset.
+// Length of data is given in dataSize.
+Status BlockAccessCipherStream::Decrypt(uint64_t fileOffset, char *data, size_t dataSize) {
+  // Calculate block index
+  auto blockSize = BlockSize();
+  uint64_t blockIndex = fileOffset / blockSize;
+  size_t blockOffset = fileOffset % blockSize;
+  unique_ptr<char[]> blockBuffer;
+
+  std::string scratch;
+  AllocateScratch(scratch);
+
+  // Decrypt individual blocks.
+  while (1) {
+    char *block = data;
+    size_t n = std::min(dataSize, blockSize - blockOffset);
+    if (n != blockSize) {
+      // We're not decrypting a full block. 
+      // Copy data to blockBuffer
+      if (!blockBuffer.get()) {
+        // Allocate buffer 
+        blockBuffer = unique_ptr<char[]>(new char[blockSize]);
+      }
+      block = blockBuffer.get();
+      // Copy encrypted data to block buffer 
+      memmove(block + blockOffset, data, n);
+    }
+    auto status = DecryptBlock(blockIndex, block, (char*)scratch.data());
+    if (!status.ok()) {
+      return status;
+    }
+    if (block != data) {
+      // Copy decrypted data back to `data`.
+      memmove(data, block + blockOffset, n);
+    }
+    dataSize -= n;
+    if (dataSize == 0) {
+      return Status::OK();
+    }
+    data += n;
+    blockOffset = 0;
+    blockIndex++;
+  }
+}
+
+// Encrypt a block of data.
+// Length of data is equal to BlockSize().
+Status ROT13BlockCipher::Encrypt(char *data) {
+  for (size_t i = 0; i < blockSize_; ++i) {
+      data[i] += 13;
+  }
+  return Status::OK();
+}
+
+// Decrypt a block of data.
+// Length of data is equal to BlockSize().
+Status ROT13BlockCipher::Decrypt(char *data) {
+  return Encrypt(data);
+}
+
+// Allocate scratch space which is passed to EncryptBlock/DecryptBlock.
+void CTRCipherStream::AllocateScratch(std::string& scratch) {
+  auto blockSize = cipher_.BlockSize();
+  scratch.reserve(blockSize);
+}
+
+// Encrypt a block of data at the given block index.
+// Length of data is equal to BlockSize();
+Status CTRCipherStream::EncryptBlock(uint64_t blockIndex, char *data, char* scratch) {
+
+  // Create nonce + counter
+  auto blockSize = cipher_.BlockSize();
+  memmove(scratch, iv_.data(), blockSize);
+  EncodeFixed64(scratch, blockIndex + initialCounter_);
+
+  // Encrypt nonce+counter 
+  auto status = cipher_.Encrypt(scratch);
+  if (!status.ok()) {
+    return status;
+  }
+
+  // XOR data with ciphertext.
+  for (size_t i = 0; i < blockSize; i++) {
+    data[i] = data[i] ^ scratch[i];
+  }
+  return Status::OK();
+}
+
+// Decrypt a block of data at the given block index.
+// Length of data is equal to BlockSize();
+Status CTRCipherStream::DecryptBlock(uint64_t blockIndex, char *data, char* scratch) {
+  // For CTR decryption & encryption are the same 
+  return EncryptBlock(blockIndex, data, scratch);
+}
+
+// GetPrefixLength returns the length of the prefix that is added to every file
+// and used for storing encryption options.
+// For optimal performance, the prefix length should be a multiple of 
+// the a page size.
+size_t CTREncryptionProvider::GetPrefixLength() {
+  return defaultPrefixLength;
+}
+
+// decodeCTRParameters decodes the initial counter & IV from the given
+// (plain text) prefix.
+static void decodeCTRParameters(const char *prefix, size_t blockSize, uint64_t &initialCounter, Slice &iv) {
+  // First block contains 64-bit initial counter
+  initialCounter = DecodeFixed64(prefix);
+  // Second block contains IV
+  iv = Slice(prefix + blockSize, blockSize);
+}
+
+// CreateNewPrefix initialized an allocated block of prefix memory 
+// for a new file.
+Status CTREncryptionProvider::CreateNewPrefix(const std::string& fname, char *prefix, size_t prefixLength) {
+  // Create & seed rnd.
+  Random rnd((uint32_t)Env::Default()->NowMicros());
+  // Fill entire prefix block with random values.
+  for (size_t i = 0; i < prefixLength; i++) {
+    prefix[i] = rnd.Uniform(256) & 0xFF;
+  }
+  // Take random data to extract initial counter & IV
+  auto blockSize = cipher_.BlockSize();
+  uint64_t initialCounter;
+  Slice prefixIV;
+  decodeCTRParameters(prefix, blockSize, initialCounter, prefixIV);
+
+  // Now populate the rest of the prefix, starting from the third block.
+  PopulateSecretPrefixPart(prefix + (2 * blockSize), prefixLength - (2 * blockSize), blockSize);
+
+  // Encrypt the prefix, starting from block 2 (leave block 0, 1 with initial counter & IV unencrypted)
+  CTRCipherStream cipherStream(cipher_, prefixIV.data(), initialCounter);
+  auto status = cipherStream.Encrypt(0, prefix + (2 * blockSize), prefixLength - (2 * blockSize));
+  if (!status.ok()) {
+    return status;
+  }
+  return Status::OK();
+}
+
+// PopulateSecretPrefixPart initializes the data into a new prefix block 
+// in plain text.
+// Returns the amount of space (starting from the start of the prefix)
+// that has been initialized.
+size_t CTREncryptionProvider::PopulateSecretPrefixPart(char *prefix, size_t prefixLength, size_t blockSize) {
+  // Nothing to do here, put in custom data in override when needed.
+  return 0;
+}
+
+Status CTREncryptionProvider::CreateCipherStream(const std::string& fname, const EnvOptions& options, Slice &prefix, unique_ptr<BlockAccessCipherStream>* result) {
+  // Read plain text part of prefix.
+  auto blockSize = cipher_.BlockSize();
+  uint64_t initialCounter;
+  Slice iv;
+  decodeCTRParameters(prefix.data(), blockSize, initialCounter, iv);
+
+  // Decrypt the encrypted part of the prefix, starting from block 2 (block 0, 1 with initial counter & IV are unencrypted)
+  CTRCipherStream cipherStream(cipher_, iv.data(), initialCounter);
+  auto status = cipherStream.Decrypt(0, (char*)prefix.data() + (2 * blockSize), prefix.size() - (2 * blockSize));
+  if (!status.ok()) {
+    return status;
+  }
+
+  // Create cipher stream 
+  return CreateCipherStreamFromPrefix(fname, options, initialCounter, iv, prefix, result);
+}
+
+// CreateCipherStreamFromPrefix creates a block access cipher stream for a file given
+// given name and options. The given prefix is already decrypted.
+Status CTREncryptionProvider::CreateCipherStreamFromPrefix(const std::string& fname, const EnvOptions& options,
+    uint64_t initialCounter, const Slice& iv, const Slice& prefix, unique_ptr<BlockAccessCipherStream>* result) {
+  (*result) = unique_ptr<BlockAccessCipherStream>(new CTRCipherStream(cipher_, iv.data(), initialCounter));
+  return Status::OK();
+}
+
+#endif // ROCKSDB_LITE
+
+}  // namespace rocksdb

env/io_posix.cc

@@ -170,9 +170,6 @@ Status PosixSequentialFile::Read(size_t n, Slice* result, char* scratch) {
       s = IOError("While reading file sequentially", filename_, errno);
     }
   }
-  // we need to fadvise away the entire range of pages because
-  // we do not want readahead pages to be cached under buffered io
-  Fadvise(fd_, 0, 0, POSIX_FADV_DONTNEED);  // free OS pages
   return s;
 }
 

env/mock_env.cc

@@ -146,7 +146,7 @@ class MemFile {
 
  private:
   uint64_t Now() {
-    int64_t unix_time;
+    int64_t unix_time = 0;
     auto s = env_->GetCurrentTime(&unix_time);
     assert(s.ok());
     return static_cast<uint64_t>(unix_time);
@@ -376,7 +376,8 @@ class TestMemLogger : public Logger {
       struct timeval now_tv;
       gettimeofday(&now_tv, nullptr);
       const time_t seconds = now_tv.tv_sec;
-      struct tm t{0, 0, 0, 0, 0, 0, 0, 0, 0};
+      struct tm t;
+      memset(&t, 0, sizeof(t));
       auto ret __attribute__((__unused__)) = localtime_r(&seconds, &t);
       assert(ret);
       p += snprintf(p, limit - p,
@@ -731,7 +732,9 @@ Status MockEnv::GetTestDirectory(std::string* path) {
 
 Status MockEnv::GetCurrentTime(int64_t* unix_time) {
   auto s = EnvWrapper::GetCurrentTime(unix_time);
-  *unix_time += fake_sleep_micros_.load() / (1000 * 1000);
+  if (s.ok()) {
+    *unix_time += fake_sleep_micros_.load() / (1000 * 1000);
+  }
   return s;
 }
 

include/rocksdb/advanced_options.h

@@ -62,6 +62,13 @@ struct CompactionOptionsFIFO {
   // Default: 1GB
   uint64_t max_table_files_size;
 
+  // Drop files older than TTL. TTL based deletion will take precedence over
+  // size based deletion if ttl > 0.
+  // delete if sst_file_creation_time < (current_time - ttl)
+  // unit: seconds. Ex: 1 day = 1 * 24 * 60 * 60
+  // Default: 0 (disabled)
+  uint64_t ttl = 0;
+
   // If true, try to do compaction to compact smaller files into larger ones.
   // Minimum files to compact follows options.level0_file_num_compaction_trigger
   // and compaction won't trigger if average compact bytes per del file is
@@ -71,9 +78,10 @@ struct CompactionOptionsFIFO {
   bool allow_compaction = false;
 
   CompactionOptionsFIFO() : max_table_files_size(1 * 1024 * 1024 * 1024) {}
-  CompactionOptionsFIFO(uint64_t _max_table_files_size,
-                        bool _allow_compaction)
+  CompactionOptionsFIFO(uint64_t _max_table_files_size, bool _allow_compaction,
+                        uint64_t _ttl = 0)
       : max_table_files_size(_max_table_files_size),
+        ttl(_ttl),
         allow_compaction(_allow_compaction) {}
 };
 
@@ -252,7 +260,7 @@ struct AdvancedColumnFamilyOptions {
   // ignore the option.
   //
   // The option is best suited for workloads where keys will likely to insert
-  // to a location close the the last inserted key with the same prefix.
+  // to a location close the last inserted key with the same prefix.
   // One example could be inserting keys of the form (prefix + timestamp),
   // and keys of the same prefix always comes in with time order. Another
   // example would be updating the same key over and over again, in which case
@@ -274,7 +282,7 @@ struct AdvancedColumnFamilyOptions {
   // If <= 0, a proper value is automatically calculated (usually 1/8 of
   // writer_buffer_size, rounded up to a multiple of 4KB).
   //
-  // There are two additional restriction of the The specified size:
+  // There are two additional restriction of the specified size:
   // (1) size should be in the range of [4096, 2 << 30] and
   // (2) be the multiple of the CPU word (which helps with the memory
   // alignment).

include/rocksdb/db.h

@@ -853,6 +853,11 @@ class DB {
     return Flush(options, DefaultColumnFamily());
   }
 
+  // Flush the WAL memory buffer to the file. If sync is true, it calls SyncWAL
+  // afterwards.
+  virtual Status FlushWAL(bool sync) {
+    return Status::NotSupported("FlushWAL not implemented");
+  }
   // Sync the wal. Note that Write() followed by SyncWAL() is not exactly the
   // same as Write() with sync=true: in the latter case the changes won't be
   // visible until the sync is done.

include/rocksdb/env.h

@@ -353,6 +353,7 @@ class Env {
   virtual Status GetHostName(char* name, uint64_t len) = 0;
 
   // Get the number of seconds since the Epoch, 1970-01-01 00:00:00 (UTC).
+  // Only overwrites *unix_time on success.
   virtual Status GetCurrentTime(int64_t* unix_time) = 0;
 
   // Get full directory name for this db.
@@ -1054,6 +1055,10 @@ class EnvWrapper : public Env {
     return target_->GetThreadID();
   }
 
+  std::string GenerateUniqueId() override {
+    return target_->GenerateUniqueId();
+  }
+
  private:
   Env* target_;
 };

include/rocksdb/env_encryption.h

@@ -0,0 +1,198 @@
+//  Copyright (c) 2016-present, Facebook, Inc.  All rights reserved.
+//  This source code is licensed under the BSD-style license found in the
+//  LICENSE file in the root directory of this source tree. An additional grant
+//  of patent rights can be found in the PATENTS file in the same directory.
+//  This source code is also licensed under the GPLv2 license found in the
+//  COPYING file in the root directory of this source tree.
+
+#pragma once
+
+#if !defined(ROCKSDB_LITE) 
+
+#include <string>
+
+#include "env.h"
+
+namespace rocksdb {
+
+class EncryptionProvider;
+
+// Returns an Env that encrypts data when stored on disk and decrypts data when 
+// read from disk.
+Env* NewEncryptedEnv(Env* base_env, EncryptionProvider* provider);
+
+// BlockAccessCipherStream is the base class for any cipher stream that 
+// supports random access at block level (without requiring data from other blocks).
+// E.g. CTR (Counter operation mode) supports this requirement.
+class BlockAccessCipherStream {
+    public:
+      virtual ~BlockAccessCipherStream() {};
+
+      // BlockSize returns the size of each block supported by this cipher stream.
+      virtual size_t BlockSize() = 0;
+
+      // Encrypt one or more (partial) blocks of data at the file offset.
+      // Length of data is given in dataSize.
+      virtual Status Encrypt(uint64_t fileOffset, char *data, size_t dataSize);
+
+      // Decrypt one or more (partial) blocks of data at the file offset.
+      // Length of data is given in dataSize.
+      virtual Status Decrypt(uint64_t fileOffset, char *data, size_t dataSize);
+
+    protected:
+      // Allocate scratch space which is passed to EncryptBlock/DecryptBlock.
+      virtual void AllocateScratch(std::string&) = 0;
+
+      // Encrypt a block of data at the given block index.
+      // Length of data is equal to BlockSize();
+      virtual Status EncryptBlock(uint64_t blockIndex, char *data, char* scratch) = 0;
+
+      // Decrypt a block of data at the given block index.
+      // Length of data is equal to BlockSize();
+      virtual Status DecryptBlock(uint64_t blockIndex, char *data, char* scratch) = 0;
+};
+
+// BlockCipher 
+class BlockCipher {
+    public:
+      virtual ~BlockCipher() {};
+
+      // BlockSize returns the size of each block supported by this cipher stream.
+      virtual size_t BlockSize() = 0;
+
+      // Encrypt a block of data.
+      // Length of data is equal to BlockSize().
+      virtual Status Encrypt(char *data) = 0;
+
+      // Decrypt a block of data.
+      // Length of data is equal to BlockSize().
+      virtual Status Decrypt(char *data) = 0;
+};
+
+// Implements a BlockCipher using ROT13.
+//
+// Note: This is a sample implementation of BlockCipher, 
+// it is NOT considered safe and should NOT be used in production.
+class ROT13BlockCipher : public BlockCipher {
+    private: 
+      size_t blockSize_;
+    public:
+      ROT13BlockCipher(size_t blockSize) 
+        : blockSize_(blockSize) {}
+      virtual ~ROT13BlockCipher() {};
+
+      // BlockSize returns the size of each block supported by this cipher stream.
+      virtual size_t BlockSize() override { return blockSize_; }
+
+      // Encrypt a block of data.
+      // Length of data is equal to BlockSize().
+      virtual Status Encrypt(char *data) override;
+
+      // Decrypt a block of data.
+      // Length of data is equal to BlockSize().
+      virtual Status Decrypt(char *data) override;
+};
+
+// CTRCipherStream implements BlockAccessCipherStream using an 
+// Counter operations mode. 
+// See https://en.wikipedia.org/wiki/Block_cipher_mode_of_operation
+//
+// Note: This is a possible implementation of BlockAccessCipherStream, 
+// it is considered suitable for use.
+class CTRCipherStream final : public BlockAccessCipherStream {
+    private:
+      BlockCipher& cipher_;
+      std::string iv_;
+      uint64_t initialCounter_;
+    public:
+      CTRCipherStream(BlockCipher& c, const char *iv, uint64_t initialCounter) 
+        : cipher_(c), iv_(iv, c.BlockSize()), initialCounter_(initialCounter) {};
+      virtual ~CTRCipherStream() {};
+
+      // BlockSize returns the size of each block supported by this cipher stream.
+      virtual size_t BlockSize() override { return cipher_.BlockSize(); }
+
+    protected:
+      // Allocate scratch space which is passed to EncryptBlock/DecryptBlock.
+      virtual void AllocateScratch(std::string&) override;
+
+      // Encrypt a block of data at the given block index.
+      // Length of data is equal to BlockSize();
+      virtual Status EncryptBlock(uint64_t blockIndex, char *data, char *scratch) override;
+
+      // Decrypt a block of data at the given block index.
+      // Length of data is equal to BlockSize();
+      virtual Status DecryptBlock(uint64_t blockIndex, char *data, char *scratch) override;
+};
+
+// The encryption provider is used to create a cipher stream for a specific file.
+// The returned cipher stream will be used for actual encryption/decryption 
+// actions.
+class EncryptionProvider {
+ public:
+    virtual ~EncryptionProvider() {};
+
+    // GetPrefixLength returns the length of the prefix that is added to every file
+    // and used for storing encryption options.
+    // For optimal performance, the prefix length should be a multiple of 
+    // the a page size.
+    virtual size_t GetPrefixLength() = 0;
+
+    // CreateNewPrefix initialized an allocated block of prefix memory 
+    // for a new file.
+    virtual Status CreateNewPrefix(const std::string& fname, char *prefix, size_t prefixLength) = 0;
+
+    // CreateCipherStream creates a block access cipher stream for a file given
+    // given name and options.
+    virtual Status CreateCipherStream(const std::string& fname, const EnvOptions& options,
+      Slice& prefix, unique_ptr<BlockAccessCipherStream>* result) = 0;
+};
+
+// This encryption provider uses a CTR cipher stream, with a given block cipher 
+// and IV.
+//
+// Note: This is a possible implementation of EncryptionProvider, 
+// it is considered suitable for use, provided a safe BlockCipher is used.
+class CTREncryptionProvider : public EncryptionProvider {
+    private:
+      BlockCipher& cipher_;
+    protected:
+      const static size_t defaultPrefixLength = 4096;
+
+ public:
+      CTREncryptionProvider(BlockCipher& c) 
+        : cipher_(c) {};
+    virtual ~CTREncryptionProvider() {}
+
+    // GetPrefixLength returns the length of the prefix that is added to every file
+    // and used for storing encryption options.
+    // For optimal performance, the prefix length should be a multiple of 
+    // the a page size.
+    virtual size_t GetPrefixLength() override;
+
+    // CreateNewPrefix initialized an allocated block of prefix memory 
+    // for a new file.
+    virtual Status CreateNewPrefix(const std::string& fname, char *prefix, size_t prefixLength) override;
+
+    // CreateCipherStream creates a block access cipher stream for a file given
+    // given name and options.
+    virtual Status CreateCipherStream(const std::string& fname, const EnvOptions& options,
+      Slice& prefix, unique_ptr<BlockAccessCipherStream>* result) override;
+
+  protected:
+    // PopulateSecretPrefixPart initializes the data into a new prefix block 
+    // that will be encrypted. This function will store the data in plain text. 
+    // It will be encrypted later (before written to disk).
+    // Returns the amount of space (starting from the start of the prefix)
+    // that has been initialized.
+    virtual size_t PopulateSecretPrefixPart(char *prefix, size_t prefixLength, size_t blockSize);
+
+    // CreateCipherStreamFromPrefix creates a block access cipher stream for a file given
+    // given name and options. The given prefix is already decrypted.
+    virtual Status CreateCipherStreamFromPrefix(const std::string& fname, const EnvOptions& options,
+      uint64_t initialCounter, const Slice& iv, const Slice& prefix, unique_ptr<BlockAccessCipherStream>* result);
+};
+
+}  // namespace rocksdb
+
+#endif  // !defined(ROCKSDB_LITE)

include/rocksdb/filter_policy.h

@@ -20,8 +20,11 @@
 #ifndef STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_
 #define STORAGE_ROCKSDB_INCLUDE_FILTER_POLICY_H_
 
-#include <string>
 #include <memory>
+#include <stdexcept>
+#include <stdlib.h>
+#include <string>
+#include <vector>
 
 namespace rocksdb {
 
@@ -41,6 +44,16 @@ class FilterBitsBuilder {
   // The return value of this function would be the filter bits,
   // The ownership of actual data is set to buf
   virtual Slice Finish(std::unique_ptr<const char[]>* buf) = 0;
+
+  // Calculate num of entries fit into a space.
+  virtual int CalculateNumEntry(const uint32_t space) {
+#ifndef ROCKSDB_LITE
+    throw std::runtime_error("CalculateNumEntry not Implemented");
+#else
+    abort();
+#endif
+    return 0;
+  }
 };
 
 // A class that checks if a key can be in filter

include/rocksdb/listener.h

@@ -71,12 +71,21 @@ enum class CompactionReason {
   kFIFOMaxSize,
   // [FIFO] reduce number of files.
   kFIFOReduceNumFiles,
+  // [FIFO] files with creation time < (current_time - interval)
+  kFIFOTtl,
   // Manual compaction
   kManualCompaction,
   // DB::SuggestCompactRange() marked files for compaction
   kFilesMarkedForCompaction,
 };
 
+enum class BackgroundErrorReason {
+  kFlush,
+  kCompaction,
+  kWriteCallback,
+  kMemTable,
+};
+
 #ifndef ROCKSDB_LITE
 
 struct TableFileDeletionInfo {
@@ -348,6 +357,20 @@ class EventListener {
   virtual void OnExternalFileIngested(
       DB* /*db*/, const ExternalFileIngestionInfo& /*info*/) {}
 
+  // A call-back function for RocksDB which will be called before setting the
+  // background error status to a non-OK value. The new background error status
+  // is provided in `bg_error` and can be modified by the callback. E.g., a
+  // callback can suppress errors by resetting it to Status::OK(), thus
+  // preventing the database from entering read-only mode. We do not provide any
+  // guarantee when failed flushes/compactions will be rescheduled if the user
+  // suppresses an error.
+  //
+  // Note that this function can run on the same threads as flush, compaction,
+  // and user writes. So, it is extremely important not to perform heavy
+  // computations or blocking calls in this function.
+  virtual void OnBackgroundError(BackgroundErrorReason /* reason */,
+                                 Status* /* bg_error */) {}
+
   // Factory method to return CompactionEventListener. If multiple listeners
   // provides CompactionEventListner, only the first one will be used.
   virtual CompactionEventListener* GetCompactionEventListener() {

include/rocksdb/metadata.h

@@ -55,17 +55,21 @@ struct LevelMetaData {
 // The metadata that describes a SST file.
 struct SstFileMetaData {
   SstFileMetaData() {}
-  SstFileMetaData(const std::string& _file_name,
-                  const std::string& _path, uint64_t _size,
-                  SequenceNumber _smallest_seqno,
+  SstFileMetaData(const std::string& _file_name, const std::string& _path,
+                  uint64_t _size, SequenceNumber _smallest_seqno,
                   SequenceNumber _largest_seqno,
                   const std::string& _smallestkey,
-                  const std::string& _largestkey,
-                  bool _being_compacted) :
-    size(_size), name(_file_name),
-    db_path(_path), smallest_seqno(_smallest_seqno), largest_seqno(_largest_seqno),
-    smallestkey(_smallestkey), largestkey(_largestkey),
-    being_compacted(_being_compacted) {}
+                  const std::string& _largestkey, uint64_t _num_reads_sampled,
+                  bool _being_compacted)
+      : size(_size),
+        name(_file_name),
+        db_path(_path),
+        smallest_seqno(_smallest_seqno),
+        largest_seqno(_largest_seqno),
+        smallestkey(_smallestkey),
+        largestkey(_largestkey),
+        num_reads_sampled(_num_reads_sampled),
+        being_compacted(_being_compacted) {}
 
   // File size in bytes.
   uint64_t size;
@@ -78,6 +82,7 @@ struct SstFileMetaData {
   SequenceNumber largest_seqno;   // Largest sequence number in file.
   std::string smallestkey;     // Smallest user defined key in the file.
   std::string largestkey;      // Largest user defined key in the file.
+  uint64_t num_reads_sampled;  // How many times the file is read.
   bool being_compacted;  // true if the file is currently being compacted.
 };
 
@@ -86,7 +91,4 @@ struct LiveFileMetaData : SstFileMetaData {
   std::string column_family_name;  // Name of the column family
   int level;               // Level at which this file resides.
 };
-
-
-
 }  // namespace rocksdb