Set version

Update readme & remove Enterprise references

.github/workflows/ci.yml

@@ -5,14 +5,17 @@ on: [push, pull_request]
 jobs:
 
   test-ubuntu-latest:
-    runs-on: ubuntu-latest
+    runs-on: self-hosted
     steps:
-    - uses: actions/checkout@v2
+    - uses: actions/checkout@v1
+      with:
+        submodules: true
     - name: make
       run: |
           sudo apt-get update
-          sudo apt-get -y install uuid-dev libcurl4-openssl-dev
-          make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' BUILD_TLS=yes -j2
+          sudo apt-get -y remove libzstd || true
+          sudo apt-get -y install uuid-dev libcurl4-openssl-dev libbz2-dev zlib1g-dev libsnappy-dev liblz4-dev libzstd-dev libgflags-dev
+          make BUILD_TLS=yes -j2 KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror'
     - name: gen-cert
       run: ./utils/gen-test-certs.sh
     - name: test-tls
@@ -28,6 +31,9 @@ jobs:
     - name: module tests
       run: |
           ./runtest-moduleapi
+    - name: rotation test
+      run: |
+          ./runtest-rotation
         
   build-ubuntu-old:
     runs-on: ubuntu-18.04
@@ -38,6 +44,7 @@ jobs:
         sudo apt-get update
         sudo apt-get -y install uuid-dev libcurl4-openssl-dev
         make -j2
+
   build-macos-latest:
     runs-on: macos-latest
     steps:
@@ -49,18 +56,4 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-    - name: make
-      run: |
-          sudo apt-get update
-          sudo apt-get -y install uuid-dev libcurl4-openssl-dev
-          make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' MALLOC=libc -j2
-  build-ubuntu-32bit:
-    runs-on: ubuntu-latest
-    steps:
-    - uses: actions/checkout@v2
-    - name: make
-      run: |
-          sudo dpkg --add-architecture i386
-          sudo apt-get update
-          sudo apt-get -y install gcc-multilib g++-multilib libc6-dev-i386 lib32z1 uuid-dev:i386 libcurl4-openssl-dev:i386
-          make KEYDB_CFLAGS='-Werror' KEYDB_CXXFLAGS='-Werror' 32bit -j2
+    - name: make

.gitlab-ci.yml

@@ -0,0 +1,164 @@
+.standard-pipeline:
+  rules:
+    - if: '$COVERAGE'
+      when: never
+    - if: '$ENDURANCE'
+      when: never
+    - if: '$CI_PIPELINE_SOURCE == "push"'
+
+build:
+  extends: .standard-pipeline
+  tags:
+    - docker
+  stage: build
+  script:
+    - git submodule update --init
+    - make distclean
+    - make -j
+  artifacts:
+    paths:
+      - src/
+
+runtest:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+  stage: test
+  script:
+    - ./runtest --config server-threads 3
+
+runtest-cluster:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+  stage: test
+  script:
+    - ./runtest-cluster
+
+runtest-moduleapi:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+  stage: test
+  script:
+    - ./runtest-moduleapi
+
+runtest-sentinel:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+  stage: test
+  script:
+    - ./runtest-sentinel
+
+node-redis-test:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  rules:
+    - when: never
+  tags:
+    - docker
+    - ipv6
+  stage: test
+  script:
+    - cp -pf src/keydb-server /usr/local/bin
+    - cp -pf src/keydb-cli /usr/local/bin
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/node-redis.git
+    - cd node-redis
+    - npm install
+    - npm run test
+
+jedis-test:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+    - ipv4
+  stage: test
+  script:
+    - cp -pf src/keydb-server /usr/local/bin
+    - cp -pf src/keydb-cli /usr/local/bin
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/jedis.git
+    - cd jedis
+    - make test
+
+redis-rs-test:
+  extends: .standard-pipeline
+  dependencies:
+    - build
+  tags:
+    - docker
+  stage: test
+  script:
+    - cp -pf src/keydb-server /usr/local/bin
+    - cp -pf src/keydb-cli /usr/local/bin
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/redis-rs.git
+    - cd redis-rs
+    - make test
+
+endurance-test:
+  rules:
+    - if: '$ENDURANCE'
+  tags:
+    - docker
+  stage: test
+  script:
+    - git submodule update --init
+    - make distclean
+    - make -j
+    - ./runtest --config server-threads 3 --loop --stop
+
+coverage-test:
+  rules:
+    - if: '$COVERAGE'
+  tags:
+    - docker
+  stage: test
+  script:
+    - git submodule update --init
+    - make distclean
+    - make gcov -j
+    - make install
+    - ./runtest || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - ./runtest-cluster || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - ./runtest-sentinel || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - ./runtest-moduleapi || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/redis-rs.git
+    - cd redis-rs
+    - make test || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - cd ..
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/jedis.git
+    - cd jedis
+    - make test || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - cd ..
+    - git clone https://gitlab-ci-token:${CI_JOB_TOKEN}@gitlab.eqalpha.com/keydb-dev/node-redis.git
+    - cd node-redis
+    - npm install
+    - npm run test || true
+    - pkill keydb-server || true
+    - pkill stunnel || true
+    - cd ..
+    - geninfo -o KeyDB.info --no-external .
+    - genhtml --legend -o lcov-html KeyDB.info

COPYING

@@ -1,7 +1,12 @@
+Code in this repository is non-free.  Portions of this code include or 
+are derivate works of code published under the BSD 3-clause license.  The
+license below applies ONLY TO THOSE PORTIONS.  Code authored by employees 
+or contractors of EQ Alpha Technology are not licensed for use without express
+written permission of EQ Alpha Technology.  All rights are reserved.
+
 Copyright (c) 2006-2020, Salvatore Sanfilippo
 Copyright (C) 2019-2021, John Sully
 Copyright (C) 2020-2021, EQ Alpha Technology Ltd.
-Copyright (C) 2022 Snap Inc.
 All rights reserved.
 
 Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

README.md

@@ -99,17 +99,11 @@ Avoid forwarding RREPLAY messages to other masters? WARNING: This setting is dan
 
 
 ```
-    db-s3-object /path/to/bucket
+    scratch-file-path /path
 ```
 If you would like KeyDB to dump and load directly to AWS S3 this option specifies the bucket.  Using this option with the traditional RDB options will result in KeyDB backing up twice to both locations.  If both are specified KeyDB will first attempt to load from the local dump file and if that fails load from S3.  This requires the AWS CLI tools to be installed and configured which are used under the hood to transfer the data.
 
 
-```
-storage-provider flash /path/to/flash
-```
-If you would like to use KeyDB FLASH storage, specify the storage medium followed by the directory path on your local SSD volume. Note that this feature is still considered experimental and should be used with discretion. See [FLASH Documentation](https://docs.keydb.dev/docs/flash) for more details on configuration and setting up your FLASH volume. 
-
-
 Building KeyDB
 --------------
 
@@ -151,12 +145,6 @@ installed):
     % ./utils/gen-test-certs.sh
     % ./runtest --tls
 
-To build with KeyDB FLASH support, use:
-
-    % make ENABLE_FLASH=yes
-
-***Note that the KeyDB FLASH feature is considered experimental (beta) and should used with discretion
-
 
 Fixing build problems with dependencies or cached build options
 ---------

build.yaml

@@ -0,0 +1,48 @@
+# Doc: https://wiki.sc-corp.net/pages/viewpage.action?pageId=121500284
+version: 1
+machamp:
+  keydb-build:
+    # Optional - build counter is linked to the build def
+    tag_template: 0.0.%build.counter%
+    # Optional - value in seconds before a build is terminated, default is 3600 seconds
+    timeout: 3600
+    # Optional - update ghe or not, default to true
+    update_ghe: true
+    code_coverage: false
+    # Required
+    steps:
+      make-build:
+        type: cmd
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./machamp_scripts/build.sh
+      tls-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest --clients $(nproc) --verbose --tls
+      cluster-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-cluster --tls
+      sentinel-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-sentinel
+      module-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-moduleapi
+      rotation-test:
+        type: cmd
+        parent: make-build
+        # https://github.sc-corp.net/Snapchat/img/tree/master/keydb/ubuntu-20-04
+        builder_image: us.gcr.io/snapchat-build-artifacts/prod/snapchat/img/keydb/keydb-ubuntu-20-04@sha256:cf869a3f5d1de1e1d976bb906689c37b7031938eb68661b844a38c532f27248c
+        command: ./runtest-rotation

ci.yaml

@@ -0,0 +1,16 @@
+# Doc: https://wiki.sc-corp.net/display/TOOL/ci.yaml+User+Guide
+version: 1
+on:
+  pull_request:
+    - workflows:
+      # All builds that use machamp should use the defined `backend_workflow`
+      - workflow_type: backend_workflow
+        # references a build defined in build.yaml
+        build_name: keydb-build
+        arch_types: ["amd64", "arm64"]
+  push:
+    - branches: [master]
+      workflows:
+        - workflow_type: backend_workflow
+          build_name: keydb-build
+          arch_types: ["amd64", "arm64"]

deps/Makefile

@@ -101,11 +101,11 @@ jemalloc: .make-prerequisites
 .PHONY: jemalloc
 
 rocksdb: .make-prerequisites
-	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
+#	@printf '%b %b\n' $(MAKECOLOR)MAKE$(ENDCOLOR) $(BINCOLOR)$@$(ENDCOLOR)
 ifeq ($(uname_M),x86_64)
-	cd rocksdb && CFLAGS=-Wno-error PORTABLE=1 USE_SSE=1 FORCE_SSE42=1 $(MAKE) static_lib
+#	cd rocksdb && CFLAGS=-Wno-error PORTABLE=1 USE_SSE=1 FORCE_SSE42=1 $(MAKE) static_lib
 else
-	cd rocksdb && PORTABLE=1 $(MAKE) static_lib
+#	cd rocksdb && PORTABLE=1 $(MAKE) static_lib
 endif
 
 .PHONY: rocksdb

keydb.conf

@@ -2071,18 +2071,4 @@ replica-weighting-factor 2
 #
 # By default this is enabled
 #
-active-client-balancing yes
-
-# Enable FLASH support (Experimental Feature)
-# storage-provider flash /path/to/flash/db
-
-# Blob support is a way to store very large objects (>200MB) on disk
-# The files are automatically cleaned up when KeyDB exits and are only
-# for temporary use.  This helps reduce memory pressure for very large
-# data items at the cost of some performance.
-#
-# By default this config is disable.  When enabled the disk associated
-# with KeyDB's working directory will be used.  If there is insufficient
-# disk space or any other I/O error KeyDB will instead use memory.
-#
-# blob-support false
+active-client-balancing yes

pkg/deb/conf/keydb.conf

@@ -2063,17 +2063,3 @@ server-threads 2
 #
 # By default KeyDB sets this to 2.
 replica-weighting-factor 2
-
-# Enable FLASH support (Experimental Feature)
-# storage-provider flash /path/to/flash/db
-
-# Blob support is a way to store very large objects (>200MB) on disk
-# The files are automatically cleaned up when KeyDB exits and are only
-# for temporary use.  This helps reduce memory pressure for very large
-# data items at the cost of some performance.
-# 
-# By default this config is disable.  When enabled the disk associated
-# with KeyDB's working directory will be used.  If there is insufficient
-# disk space or any other I/O error KeyDB will instead use memory.
-# 
-# blob-support false

pkg/deb/deb-buildsource.sh

@@ -19,17 +19,12 @@ elif [ "$distributor" == "Ubuntu" ]; then
 fi
 codename=$(lsb_release --codename --short)
 date=$(date +%a," "%d" "%b" "%Y" "%T)
-
-# overwrite debian bookworm version until updated
-if [ $codename == "bookworm" ]; then
-    distname=+deb12u1
-fi
 pkg_name=keydb-$majorv:$version$distname
 
 # create build tree
 cd ../../../
-tar -czvf keydb_$version.orig.tar.gz --force-local KeyDB
-cd KeyDB/pkg/deb/
+tar -czvf keydb_$version.orig.tar.gz --force-local keydb-internal
+cd keydb-internal/pkg/deb/
 mkdir -p $pkg_name/tmp
 if [[ "$codename" == "xenial" ]] || [[ "$codename" == "stretch" ]]; then
 	cp -r debian_dh9 $pkg_name/tmp/debian

pkg/deb/debian/control

@@ -26,8 +26,8 @@ Build-Depends:
  libzstd-dev <!nocheck>
 Standards-Version: 4.2.1
 Homepage: https://docs.keydb.dev/
-Vcs-Git: https://github.com/Snapchat/KeyDB.git
-Vcs-Browser: https://github.com/Snapchat/KeyDB
+Vcs-Git: https://github.com/EQ-Alpha/KeyDB.git
+Vcs-Browser: https://github.com/EQ-Alpha/KeyDB
 
 Package: keydb
 Architecture: all

pkg/deb/debian/rules

@@ -4,7 +4,6 @@ include /usr/share/dpkg/buildflags.mk
 
 export BUILD_TLS=yes
 export USE_SYSTEMD=yes
-export ENABLE_FLASH=yes
 export CFLAGS CPPFLAGS LDFLAGS
 export DEB_BUILD_MAINT_OPTIONS = hardening=+all
 export DEB_LDFLAGS_MAINT_APPEND = -ldl -latomic $(LUA_LDFLAGS)

pkg/deb/debian_dh9/control

@@ -25,8 +25,8 @@ Build-Depends:
  libzstd-dev <!nocheck>
 Standards-Version: 4.2.1
 Homepage: https://docs.keydb.dev/
-Vcs-Git: https://github.com/Snapchat/KeyDB.git
-Vcs-Browser: https://github.com/Snapchat/KeyDB
+Vcs-Git: https://github.com/EQ-Alpha/KeyDB.git
+Vcs-Browser: https://github.com/EQ-Alpha/KeyDB
 
 Package: keydb
 Architecture: all

pkg/deb/debian_dh9/rules

@@ -10,7 +10,6 @@ include /usr/share/dpkg/buildflags.mk
 
 export BUILD_TLS=yes
 export USE_SYSTEMD=yes
-export ENABLE_FLASH=yes
 export CFLAGS CPPFLAGS LDFLAGS
 export DEB_BUILD_MAINT_OPTIONS = hardening=+all
 export DEB_LDFLAGS_MAINT_APPEND = -ldl -latomic $(LUA_LDFLAGS)

pkg/deb/master_changelog

@@ -1,25 +1,3 @@
-keydb (6:6.3.1-1distribution_placeholder) codename_placeholder; urgency=medium
-
-  * This point release contains fixes to bugs related to expires, active-rep, and rdb saving
-  * Issues fixed: #419, #422, #428
-  * PRs: #426, #429, #431, #433
-
--- Ben Schermel <ben@eqalpha.com>  Mon, 23 May 2022 20:00:37 +0000
-
-keydb (6:6.3.0-1distribution_placeholder) codename_placeholder; urgency=medium
-
-  * This release open sources KeyDB Enterprise features into the open source project along with PSYNC for active replication
-  * Partial synchronization for active replication is introduced
-  * MVCC introduced into codebase from KeyDB Enterprise
-  * Async commands added: GET, MGET. These will see perf improvements
-  * KEYS and SCAN commands will no longer be blocking calls
-  * Async Rehash implemented for additional stability to perf
-  * IStorage interface added
-  * In-process background saving (forkless) to comply with maxmemory setting
-  * See v6.3.0 tagged release notes on github for a detailed explanation of these changes
-
--- Ben Schermel <ben@eqalpha.com>  Wed, 11 May 2022 20:00:37 +0000
-
 keydb (6:6.2.2-1distribution_placeholder) codename_placeholder; urgency=medium
 
   * Acquire lock in module.cpp to fix module test break

pkg/docker/Dockerfile

@@ -1,4 +1,4 @@
-FROM ubuntu:18.04
+FROM ubuntu:20.04
 SHELL ["/bin/bash","-c"]
 RUN groupadd -r keydb && useradd -r -g keydb keydb
 # use gosu for easy step-down from root: https://github.com/tianon/gosu/releases
@@ -77,7 +77,7 @@ RUN \
         mkdir /data && chown keydb:keydb /data; \
         mkdir /flash && chown keydb:keydb /flash; \
         mkdir -p /etc/keydb; \
-        cp /tmp/KeyDB/keydb.conf /etc/keydb/; \
+        cp /tmp/keydb-internal/keydb.conf /etc/keydb/; \
         sed -i 's/^\(daemonize .*\)$/# \1/' /etc/keydb/keydb.conf; \
         sed -i 's/^\(dir .*\)$/# \1\ndir \/data/' /etc/keydb/keydb.conf; \
         sed -i 's/^\(logfile .*\)$/# \1/' /etc/keydb/keydb.conf; \
@@ -110,3 +110,8 @@ ENV KEYDB_PRO_DIRECTORY=/usr/local/bin/
 ENTRYPOINT ["docker-entrypoint.sh"]
 EXPOSE 6379
 CMD ["keydb-server","/etc/keydb/keydb.conf"]
+
+
+
+
+

pkg/docker/Dockerfile_Alpine

@@ -1,77 +0,0 @@
-FROM alpine:3.12
-# add our user and group first to make sure their IDs get assigned consistently, regardless of whatever dependencies get added
-RUN addgroup -S -g 1000 keydb && adduser -S -G keydb -u 999 keydb
-RUN mkdir -p /etc/keydb
-ARG BRANCH
-RUN set -eux; \
-        \
-        apk add --no-cache su-exec; \
-        apk add --no-cache --virtual .build-deps \
-                coreutils \
-                gcc \
-                linux-headers \
-                make \
-                musl-dev \
-                openssl-dev \
-                git \
-                util-linux-dev \
-                curl-dev \
-                g++ \
-                libunwind-dev \
-                bash \
-                perl \
-                git \
-        ; \
-        cd /tmp && git clone --branch $BRANCH https://github.com/Snapchat/KeyDB.git --recursive; \
-        cd /tmp/KeyDB; \
-        # disable protected mode as it relates to docker
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *1 *,.*[)],$' ./src/config.cpp; \
-        sed -ri 's!^( *createBoolConfig[(]"protected-mode",.*, *)1( *,.*[)],)$!\10\2!' ./src/config.cpp; \
-        grep -E '^ *createBoolConfig[(]"protected-mode",.*, *0 *,.*[)],$' ./src/config.cpp; \
-        make -j$(nproc) BUILD_TLS=yes; \
-        cd src; \
-        strip keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel keydb-server; \
-        mv keydb-server keydb-cli keydb-benchmark keydb-check-rdb keydb-check-aof keydb-diagnostic-tool keydb-sentinel /usr/local/bin/; \
-        runDeps="$( \
-                scanelf --needed --nobanner --format '%n#p' --recursive /usr/local \
-                        | tr ',' '\n' \
-                        | sort -u \
-                        | awk 'system("[ -e /usr/local/lib/" $1 " ]") == 0 { next } { print "so:" $1 }' \
-        )"; \
-        apk add --no-network --virtual .keydb-rundeps $runDeps; \
-        apk del --no-network .build-deps; \
-        # create working directories and organize files
-        mkdir /data && chown keydb:keydb /data; \
-        mkdir /flash && chown keydb:keydb /flash; \
-        mkdir -p /etc/keydb; \
-        cp /tmp/KeyDB/keydb.conf /etc/keydb/; \
-        sed -i 's/^\(daemonize .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(dir .*\)$/# \1\ndir \/data/' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(logfile .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        sed -i 's/protected-mode yes/protected-mode no/g' /etc/keydb/keydb.conf; \
-        sed -i 's/^\(bind .*\)$/# \1/' /etc/keydb/keydb.conf; \
-        ln -s keydb-cli redis-cli; \
-        cd /etc/keydb; \
-        ln -s keydb.conf redis.conf; \
-        rm -rf /tmp/*
-# generate entrypoint script
-RUN set -eux; \
-        echo '#!/bin/sh' > /usr/local/bin/docker-entrypoint.sh; \
-        echo 'set -e' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# first arg is '-f' or '--some-option'" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# or first arg is `something.conf`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "${1#-}" != "$1" ] || [ "${1%.conf}" != "$1" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        set -- keydb-server "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "# allow the container to be started with `--user`" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'if [ "$1" = "keydb-server" -a "$(id -u)" = "0" ]; then' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo "        find . \! -user keydb -exec chown keydb '{}' +" >> /usr/local/bin/docker-entrypoint.sh; \
-        echo '        exec su-exec keydb "$0" "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'fi' >> /usr/local/bin/docker-entrypoint.sh; \
-        echo 'exec "$@"' >> /usr/local/bin/docker-entrypoint.sh; \
-        chmod +x /usr/local/bin/docker-entrypoint.sh
-VOLUME /data
-WORKDIR /data
-ENTRYPOINT ["docker-entrypoint.sh"]
-EXPOSE 6379
-CMD ["keydb-server", "/etc/keydb/keydb.conf"]

pkg/rpm/keydb_build/keydb_rpm/etc/keydb/keydb.conf

@@ -2063,17 +2063,3 @@ server-threads 2
 #
 # By default KeyDB sets this to 2.
 replica-weighting-factor 2
-
-# Enable FLASH support (Experimental Feature)
-# storage-provider flash /path/to/flash/db
-
-# Blob support is a way to store very large objects (>200MB) on disk
-# The files are automatically cleaned up when KeyDB exits and are only
-# for temporary use.  This helps reduce memory pressure for very large
-# data items at the cost of some performance.
-# 
-# By default this config is disable.  When enabled the disk associated
-# with KeyDB's working directory will be used.  If there is insufficient
-# disk space or any other I/O error KeyDB will instead use memory.
-# 
-# blob-support false

src/Makefile

@@ -20,7 +20,7 @@ release_hdr := $(shell sh -c './mkreleasehdr.sh')
 uname_S := $(shell sh -c 'uname -s 2>/dev/null || echo not')
 uname_M := $(shell sh -c 'uname -m 2>/dev/null || echo not')
 OPTIMIZATION?=-O2 -flto
-DEPENDENCY_TARGETS=hiredis linenoise lua hdr_histogram
+DEPENDENCY_TARGETS=hiredis linenoise lua hdr_histogram rocksdb
 NODEPS:=clean distclean
 
 # Default settings
@@ -62,7 +62,6 @@ endif
 endif
 
 USEASM?=true
-ENABLE_FLASH?=no
 
 ifneq ($(strip $(SANITIZE)),)
 	CFLAGS+= -fsanitize=$(SANITIZE) -DSANITIZE  -fno-omit-frame-pointer
@@ -72,15 +71,6 @@ ifneq ($(strip $(SANITIZE)),)
 	USEASM=false
 endif
 
-ifeq ($(ENABLE_FLASH),yes)
-	FINAL_LIBS+= -lz -lcrypto -lbz2 -lzstd -llz4 -lsnappy
-	CXXFLAGS+= -I../deps/rocksdb/include/ -DENABLE_ROCKSDB
-	STORAGE_OBJ+= storage/rocksdb.o storage/rocksdbfactory.o
-	FINAL_LIBS+= ../deps/rocksdb/librocksdb.a 
-	DEPENDENCY_TARGETS+= rocksdb
-endif
-
-
 ifeq ($(CHECKED),true)
 	CXXFLAGS+= -DCHECKED_BUILD
 endif
@@ -142,7 +132,7 @@ DEBUG=-g -ggdb
 FINAL_CFLAGS=$(STD) $(WARN) $(OPT) $(DEBUG) $(CFLAGS) $(KEYDB_CFLAGS) $(REDIS_CFLAGS)
 FINAL_CXXFLAGS=$(CXX_STD) $(WARN) $(OPT) $(DEBUG) $(CXXFLAGS) $(KEYDB_CFLAGS) $(REDIS_CFLAGS)
 FINAL_LDFLAGS=$(LDFLAGS) $(KEYDB_LDFLAGS) $(DEBUG)
-FINAL_LIBS+=-lm
+FINAL_LIBS+=-lm -lz -lcrypto
 
 ifneq ($(uname_S),Darwin)
 	FINAL_LIBS+=-latomic
@@ -260,7 +250,7 @@ endif
 
 # Include paths to dependencies
 FINAL_CFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram
-FINAL_CXXFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram -I../deps/concurrentqueue
+FINAL_CXXFLAGS+= -I../deps/hiredis -I../deps/linenoise -I../deps/lua/src -I../deps/hdr_histogram -I../deps/rocksdb/include/ -I../deps/concurrentqueue
 
 # Determine systemd support and/or build preference (defaulting to auto-detection)
 BUILD_WITH_SYSTEMD=no
@@ -376,7 +366,7 @@ endif
 
 REDIS_SERVER_NAME=keydb-server$(PROG_SUFFIX)
 REDIS_SENTINEL_NAME=keydb-sentinel$(PROG_SUFFIX)
-REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o t_nhash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crcspeed.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.o localtime.o acl.o storage.o rdb-s3.o fastlock.o new.o tracking.o cron.o connection.o tls.o sha256.o motd_server.o timeout.o setcpuaffinity.o AsyncWorkQueue.o snapshot.o storage/teststorageprovider.o keydbutils.o StorageCache.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ) $(STORAGE_OBJ)
+REDIS_SERVER_OBJ=adlist.o quicklist.o ae.o anet.o dict.o server.o sds.o zmalloc.o lzf_c.o lzf_d.o pqsort.o zipmap.o sha1.o ziplist.o release.o networking.o util.o object.o db.o replication.o rdb.o t_string.o t_list.o t_set.o t_zset.o t_hash.o t_nhash.o config.o aof.o pubsub.o multi.o debug.o sort.o intset.o syncio.o cluster.o crc16.o endianconv.o slowlog.o scripting.o bio.o rio.o rand.o memtest.o crcspeed.o crc64.o bitops.o sentinel.o notify.o setproctitle.o blocked.o hyperloglog.o latency.o sparkline.o redis-check-rdb.o redis-check-aof.o geo.o lazyfree.o module.o evict.o expire.o geohash.o geohash_helper.o childinfo.o defrag.o siphash.o rax.o t_stream.o listpack.o localtime.o acl.o storage.o rdb-s3.o fastlock.o new.o tracking.o cron.o connection.o tls.o sha256.o motd_server.o timeout.o setcpuaffinity.o AsyncWorkQueue.o snapshot.o storage/teststorageprovider.o keydbutils.o StorageCache.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)
 KEYDB_SERVER_OBJ=SnapshotPayloadParseState.o
 REDIS_CLI_NAME=keydb-cli$(PROG_SUFFIX)
 REDIS_CLI_OBJ=anet.o adlist.o dict.o redis-cli.o redis-cli-cpphelper.o zmalloc.o release.o anet.o ae.o crcspeed.o crc64.o siphash.o crc16.o storage-lite.o fastlock.o motd_client.o monotonic.o cli_common.o mt19937-64.o $(ASM_OBJ)

src/StorageCache.cpp

@@ -203,7 +203,6 @@ void StorageCache::beginWriteBatch() {
 }
 
 void StorageCache::emergencyFreeCache() {
-    std::unique_lock<fastlock> ul(m_lock);
     dict *d = m_pdict;
     m_pdict = nullptr;
     if (d != nullptr) {

src/cluster.cpp

@@ -771,7 +771,7 @@ unsigned long getClusterConnectionsCount(void) {
     /* We decrement the number of nodes by one, since there is the
      * "myself" node too in the list. Each node uses two file descriptors,
      * one incoming and one outgoing, thus the multiplication by 2. */
-    return g_pserver->cluster_enabled && g_pserver->cluster != nullptr ?
+    return g_pserver->cluster_enabled ?
            ((dictSize(g_pserver->cluster->nodes)-1)*2) : 0;
 }
 
@@ -1525,9 +1525,7 @@ void clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) {
                  * it's greater than our view but is not in the future
                  * (with 500 milliseconds tolerance) from the POV of our
                  * clock. */
-                mstime_t mstime;
-                __atomic_load(&g_pserver->mstime, &mstime, __ATOMIC_RELAXED);
-                if (pongtime <= (mstime+500) &&
+                if (pongtime <= (g_pserver->mstime+500) &&
                     pongtime > node->pong_received)
                 {
                     node->pong_received = pongtime;

src/config.cpp

@@ -355,17 +355,16 @@ bool initializeStorageProvider(const char **err)
             return true;
         if (!strcasecmp(g_sdsProvider, "flash") && g_sdsArgs != nullptr)
         {
-#ifdef ENABLE_ROCKSDB
+            // Temporary: Disable FLASH
+            serverLog(LL_WARNING, "FLASH Is Not yet Supported");
+            _Exit(EXIT_FAILURE);
+
+
             // Create The Storage Factory (if necessary)
             serverLog(LL_NOTICE, "Initializing FLASH storage provider (this may take a long time)");
             adjustOpenFilesLimit();
             g_pserver->m_pstorageFactory = CreateRocksDBStorageFactory(g_sdsArgs, cserver.dbnum, cserver.storage_conf, cserver.storage_conf ? strlen(cserver.storage_conf) : 0);
-#else
-	    serverLog(LL_WARNING, "To use the flash storage provider please compile KeyDB with ENABLE_FLASH=yes");
-	    serverLog(LL_WARNING, "Exiting due to the use of an unsupported storage provider");
-	    exit(EXIT_FAILURE);
-#endif
-	}
+        }
         else if (!strcasecmp(g_sdsProvider, "test") && g_sdsArgs == nullptr)
         {
             g_pserver->m_pstorageFactory = new (MALLOC_LOCAL) TestStorageFactory();
@@ -767,12 +766,6 @@ void loadServerConfigFromString(char *config) {
             g_sdsProvider = sdsdup(argv[1]);
             if (argc > 2)
                 g_sdsArgs = sdsdup(argv[2]);
-	} else if (!strcasecmp(argv[0],"is-flash-enabled") && argc == 1) {
-#ifdef ENABLE_ROCKSDB
-            exit(EXIT_SUCCESS);
-#else
-            exit(EXIT_FAILURE);
-#endif
         } else {
             err = "Bad directive or wrong number of arguments"; goto loaderr;
         }
@@ -2789,7 +2782,7 @@ standardConfig configs[] = {
     createBoolConfig("disable-thp", NULL, MODIFIABLE_CONFIG, g_pserver->disable_thp, 1, NULL, NULL),
     createBoolConfig("cluster-allow-replica-migration", NULL, MODIFIABLE_CONFIG, g_pserver->cluster_allow_replica_migration, 1, NULL, NULL),
     createBoolConfig("replica-announced", NULL, MODIFIABLE_CONFIG, g_pserver->replica_announced, 1, NULL, NULL),
-    createBoolConfig("enable-async-commands", NULL, MODIFIABLE_CONFIG, g_pserver->enable_async_commands, 0, NULL, NULL),
+    createBoolConfig("enable-async-commands", NULL, MODIFIABLE_CONFIG, g_pserver->enable_async_commands, 1, NULL, NULL),
     createBoolConfig("multithread-load-enabled", NULL, MODIFIABLE_CONFIG, g_pserver->multithread_load_enabled, 0, NULL, NULL),
     createBoolConfig("active-client-balancing", NULL, MODIFIABLE_CONFIG, g_pserver->active_client_balancing, 1, NULL, NULL),
 
@@ -2909,7 +2902,6 @@ standardConfig configs[] = {
     createBoolConfig("multi-master-no-forward", NULL, MODIFIABLE_CONFIG, cserver.multimaster_no_forward, 0, validateMultiMasterNoForward, NULL),
     createBoolConfig("allow-write-during-load", NULL, MODIFIABLE_CONFIG, g_pserver->fWriteDuringActiveLoad, 0, NULL, NULL),
     createBoolConfig("force-backlog-disk-reserve", NULL, MODIFIABLE_CONFIG, cserver.force_backlog_disk, 0, NULL, NULL),
-    createBoolConfig("soft-shutdown", NULL, MODIFIABLE_CONFIG, g_pserver->config_soft_shutdown, 0, NULL, NULL),
 
 #ifdef USE_OPENSSL
     createIntConfig("tls-port", NULL, MODIFIABLE_CONFIG, 0, 65535, g_pserver->tls_port, 0, INTEGER_CONFIG, NULL, updateTLSPort), /* TCP port. */

src/db.cpp

@@ -1462,11 +1462,6 @@ void shutdownCommand(client *c) {
             flags |= SHUTDOWN_NOSAVE;
         } else if (!strcasecmp(szFromObj(c->argv[1]),"save")) {
             flags |= SHUTDOWN_SAVE;
-        } else if (!strcasecmp(szFromObj(c->argv[1]), "soft")) {
-            g_pserver->soft_shutdown = true;
-            serverLog(LL_WARNING, "Soft Shutdown Initiated");
-            addReply(c, shared.ok);
-            return;
         } else {
             addReplyErrorObject(c,shared.syntaxerr);
             return;
@@ -1573,12 +1568,6 @@ void moveCommand(client *c) {
         return;
     }
 
-    /* Return zero if the key already exists in the target DB */
-    if (lookupKeyWrite(dst,c->argv[1]) != NULL) {
-        addReply(c,shared.czero);
-        return;
-    }
-
     std::unique_ptr<expireEntry> spexpire;
     {   // scope pexpireOld
     std::unique_lock<fastlock> ul(g_expireLock);
@@ -1593,6 +1582,12 @@ void moveCommand(client *c) {
     dbDelete(src,c->argv[1]);
     g_pserver->dirty++;
 
+    /* Return zero if the key already exists in the target DB */
+    if (lookupKeyWrite(dst,c->argv[1]) != NULL) {
+        addReply(c,shared.czero);
+        decrRefCount(o);
+        return;
+    }
     dbAdd(dst,c->argv[1],o);
     if (spexpire != nullptr) setExpire(c,dst,c->argv[1],std::move(*spexpire));
 
@@ -2644,10 +2639,7 @@ bool redisDbPersistentData::insert(char *key, robj *o, bool fAssumeNew, dict_ite
         ensure(key);
     dictEntry *de;
     int res = dictAdd(m_pdict, key, o, &de);
-    if (!FImplies(fAssumeNew, res == DICT_OK)) {
-        serverLog(LL_WARNING,
-            "Assumed new key %s existed in DB.", key);
-    }
+    serverAssert(FImplies(fAssumeNew, res == DICT_OK));
     if (res == DICT_OK)
     {
 #ifdef CHECKED_BUILD
@@ -2677,11 +2669,6 @@ void redisDbPersistentData::prepOverwriteForSnapshot(char *key)
         auto itr = m_pdbSnapshot->find_cached_threadsafe(key);
         if (itr.key() != nullptr)
         {
-            if (itr.val()->FExpires()) {
-                // Note: I'm sure we could handle this, but its too risky at the moment.
-                //  There are known bugs doing this with expires
-                return;
-            }
             sds keyNew = sdsdupshared(itr.key());
             if (dictAdd(m_pdictTombstone, keyNew, (void*)dictHashKey(m_pdict, key)) != DICT_OK)
                 sdsfree(keyNew);
@@ -3276,11 +3263,10 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
                     dictEntry **table;
                     __atomic_load(&c->db->m_pdict->ht[iht].table, &table, __ATOMIC_RELAXED);
                     if (table != nullptr) {
-                        dictEntry *de;
-                        __atomic_load(&table[hT], &de, __ATOMIC_ACQUIRE);
+                        dictEntry *de = table[hT];
                         while (de != nullptr) {
                             _mm_prefetch(dictGetKey(de), _MM_HINT_T2);
-                            __atomic_load(&de->next, &de, __ATOMIC_ACQUIRE);
+                            de = de->next;
                         }
                     }
                     if (!dictIsRehashing(c->db->m_pdict))
@@ -3300,10 +3286,6 @@ bool redisDbPersistentData::prefetchKeysAsync(client *c, parsed_command &command
     auto cmd = lookupCommand(szFromObj(command.argv[0]));
     if (cmd == nullptr)
         return false; // Bad command? It's not for us to judge, just bail
-    
-    if (command.argc < std::abs(cmd->arity))
-        return false; // Invalid number of args
-    
     int numkeys = getKeysFromCommand(cmd, command.argv, command.argc, &result);
     for (int ikey = 0; ikey < numkeys; ++ikey)
     {

src/dict.cpp

@@ -128,7 +128,6 @@ int _dictInit(dict *d, dictType *type,
     d->pauserehash = 0;
     d->asyncdata = nullptr;
     d->refcount = 1;
-    d->noshrink = false;
     return DICT_OK;
 }
 
@@ -205,7 +204,7 @@ int dictMerge(dict *dst, dict *src)
     
     if (dictSize(dst) == 0)
     {
-        dict::swap(*dst, *src);
+        std::swap(*dst, *src);
         std::swap(dst->pauserehash, src->pauserehash);
         return DICT_OK;
     }
@@ -213,7 +212,7 @@ int dictMerge(dict *dst, dict *src)
     size_t expectedSize = dictSize(src) + dictSize(dst);
     if (dictSize(src) > dictSize(dst) && src->asyncdata == nullptr && dst->asyncdata == nullptr)
     {
-        dict::swap(*dst, *src);
+        std::swap(*dst, *src);
         std::swap(dst->pauserehash, src->pauserehash);
     }
 
@@ -403,7 +402,7 @@ int dictRehash(dict *d, int n) {
 
 dictAsyncRehashCtl::dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next) : dict(d), next(next) {
     queue.reserve(c_targetQueueSize);
-    __atomic_fetch_add(&d->refcount, 1, __ATOMIC_ACQ_REL);
+    __atomic_fetch_add(&d->refcount, 1, __ATOMIC_RELEASE);
     this->rehashIdxBase = d->rehashidx;
 }
 
@@ -447,9 +446,6 @@ LDone:
 }
 
 void dictRehashAsync(dictAsyncRehashCtl *ctl) {
-    if (ctl->abondon.load(std::memory_order_acquire)) {
-        ctl->hashIdx = ctl->queue.size();
-    }
     for (size_t idx = ctl->hashIdx; idx < ctl->queue.size(); ++idx) {
         auto &wi = ctl->queue[idx];
         wi.hash = dictHashKey(ctl->dict, dictGetKey(wi.de));
@@ -459,9 +455,6 @@ void dictRehashAsync(dictAsyncRehashCtl *ctl) {
 }
 
 bool dictRehashSomeAsync(dictAsyncRehashCtl *ctl, size_t hashes) {
-    if (ctl->abondon.load(std::memory_order_acquire)) {
-        ctl->hashIdx = ctl->queue.size();
-    }
     size_t max = std::min(ctl->hashIdx + hashes, ctl->queue.size());
     for (; ctl->hashIdx < max; ++ctl->hashIdx) {
         auto &wi = ctl->queue[ctl->hashIdx];
@@ -472,23 +465,6 @@ bool dictRehashSomeAsync(dictAsyncRehashCtl *ctl, size_t hashes) {
     return ctl->hashIdx < ctl->queue.size();
 }
 
-
-void discontinueAsyncRehash(dict *d) {
-    // We inform our async rehashers and the completion function the results are to be
-    //  abandoned.  We keep the asyncdata linked in so that dictEntry's are still added
-    //  to the GC list.  This is because we can't gurantee when the other threads will
-    //  stop looking at them.
-    if (d->asyncdata != nullptr) {
-        auto adata = d->asyncdata;
-        while (adata != nullptr && !adata->abondon.load(std::memory_order_relaxed)) {
-            adata->abondon = true;
-            adata = adata->next;
-        }
-        if (dictIsRehashing(d))
-            d->rehashidx = 0;
-    }
-}
-
 void dictCompleteRehashAsync(dictAsyncRehashCtl *ctl, bool fFree) {
     dict *d = ctl->dict;
     assert(ctl->done);
@@ -810,8 +786,6 @@ int _dictClear(dict *d, dictht *ht, void(callback)(void *)) {
         if (callback && (i & 65535) == 0) callback(d->privdata);
 
         if ((he = ht->table[i]) == NULL) continue;
-        dictEntry *deNull = nullptr;
-        __atomic_store(&ht->table[i], &deNull, __ATOMIC_RELEASE);
         while(he) {
             nextHe = he->next;
             if (d->asyncdata && (ssize_t)i < d->rehashidx) {

src/dict.h

@@ -110,16 +110,12 @@ struct dictAsyncRehashCtl {
     std::atomic<bool> abondon { false };
 
     dictAsyncRehashCtl(struct dict *d, dictAsyncRehashCtl *next);
-    dictAsyncRehashCtl(const dictAsyncRehashCtl&) = delete;
-    dictAsyncRehashCtl(dictAsyncRehashCtl&&) = delete;
     ~dictAsyncRehashCtl();
 };
 #else
 struct  dictAsyncRehashCtl;
 #endif
 
-void discontinueAsyncRehash(dict *d);
-
 typedef struct dict {
     dictType *type;
     void *privdata;
@@ -129,24 +125,6 @@ typedef struct dict {
     dictAsyncRehashCtl *asyncdata;
     int16_t pauserehash; /* If >0 rehashing is paused (<0 indicates coding error) */
     uint8_t noshrink = false;
-
-#ifdef __cplusplus
-    dict() = default;
-    dict(dict &) = delete;  // No Copy Ctor
-
-    static void swap(dict& a, dict& b) {
-        discontinueAsyncRehash(&a);
-        discontinueAsyncRehash(&b);
-        std::swap(a.type, b.type);
-        std::swap(a.privdata, b.privdata);
-        std::swap(a.ht[0], b.ht[0]);
-        std::swap(a.ht[1], b.ht[1]);
-        std::swap(a.rehashidx, b.rehashidx);
-        // Never swap refcount - they are attached to the specific dict obj
-        std::swap(a.pauserehash, b.pauserehash);
-        std::swap(a.noshrink, b.noshrink);
-    }
-#endif
 } dict;
 
 /* If safe is set to 1 this is a safe iterator, that means, you can call

src/gc.h

@@ -2,7 +2,6 @@
 #include <vector>
 #include <assert.h>
 #include <unordered_set>
-#include <list>
 
 struct ICollectable
 {
@@ -16,17 +15,7 @@ class GarbageCollector
     struct EpochHolder
     {
         uint64_t tstamp;
-        std::unique_ptr<std::vector<std::unique_ptr<T>>> m_spvecObjs;
-
-        EpochHolder() {
-            m_spvecObjs = std::make_unique<std::vector<std::unique_ptr<T>>>();
-        }
-
-        // Support move operators
-        EpochHolder(EpochHolder &&other) = default;
-        EpochHolder &operator=(EpochHolder &&) = default;
-
-
+        std::vector<std::unique_ptr<T>> m_vecObjs;
 
         bool operator<(uint64_t tstamp) const
         {
@@ -56,14 +45,14 @@ public:
     void shutdown()
     {
         std::unique_lock<fastlock> lock(m_lock);
-        m_listepochs.clear();
+        m_vecepochs.clear();
         m_setepochOutstanding.clear();
     }
 
     bool empty() const
     {
         std::unique_lock<fastlock> lock(m_lock);
-        return m_listepochs.empty();
+        return m_vecepochs.empty();
     }
 
     void endEpoch(uint64_t epoch, bool fNoFree = false)
@@ -74,12 +63,12 @@ public:
         m_setepochOutstanding.erase(epoch);
         if (fNoFree)
             return;
-        std::list<EpochHolder> listclean;
+        std::vector<EpochHolder> vecclean;
         
         // No outstanding epochs?
         if (m_setepochOutstanding.empty())
         {
-            listclean = std::move(m_listepochs);  // Everything goes!
+            vecclean = std::move(m_vecepochs);  // Everything goes!
         }
         else
         {
@@ -88,20 +77,18 @@ public:
                 return; // No available epochs to free
 
             // Clean any epochs available (after the lock)
-            for (auto itr = m_listepochs.begin(); itr != m_listepochs.end(); /* itr incremented in loop*/)
+            for (size_t iepoch = 0; iepoch < m_vecepochs.size(); ++iepoch)
             {
-                auto &e = *itr;
-                auto itrNext = itr;
-                ++itrNext;
+                auto &e = m_vecepochs[iepoch];
                 if (e < minepoch)
                 {
-                    listclean.emplace_back(std::move(e));
-                    m_listepochs.erase(itr);
+                    vecclean.emplace_back(std::move(e));
+                    m_vecepochs.erase(m_vecepochs.begin() + iepoch);
+                    --iepoch;
                 }
-                itr = itrNext;
             }
 
-            assert(listclean.empty() || fMinElement);
+            assert(vecclean.empty() || fMinElement);
         }
 
         lock.unlock();  // don't hold it for the potentially long delete of vecclean
@@ -113,24 +100,24 @@ public:
         serverAssert(m_setepochOutstanding.find(epoch) != m_setepochOutstanding.end());
         serverAssert(sp->FWillFreeChildDebug() == false);
 
-        auto itr = std::find(m_listepochs.begin(), m_listepochs.end(), m_epochNext+1);
-        if (itr == m_listepochs.end())
+        auto itr = std::find(m_vecepochs.begin(), m_vecepochs.end(), m_epochNext+1);
+        if (itr == m_vecepochs.end())
         {
             EpochHolder e;
             e.tstamp = m_epochNext+1;
-            e.m_spvecObjs->push_back(std::move(sp));
-            m_listepochs.emplace_back(std::move(e));
+            e.m_vecObjs.push_back(std::move(sp));
+            m_vecepochs.emplace_back(std::move(e));
         }
         else
         {
-            itr->m_spvecObjs->push_back(std::move(sp));
+            itr->m_vecObjs.push_back(std::move(sp));
         }
     }
 
 private:
     mutable fastlock m_lock { "Garbage Collector"};
 
-    std::list<EpochHolder> m_listepochs;
+    std::vector<EpochHolder> m_vecepochs;
     std::unordered_set<uint64_t> m_setepochOutstanding;
     uint64_t m_epochNext = 0;
-};
+};

src/help.h

@@ -1071,7 +1071,7 @@ struct commandHelp {
     1,
     "2.2.0" },
     { "SHUTDOWN",
-    "[NOSAVE|SAVE|SOFT]",
+    "[NOSAVE|SAVE]",
     "Synchronously save the dataset to disk and then shut down the server",
     9,
     "1.0.0" },

src/networking.cpp

@@ -1256,20 +1256,6 @@ static void acceptCommonHandler(connection *conn, int flags, char *ip, int iel)
         return;
     }
 
-    /* Prevent new connections if we're in a soft shutdown situation */
-    if (g_pserver->soft_shutdown) {
-        const char *err = "-SHUTDOWN";
-        /* That's a best effort error message, don't check write errors.
-         * Note that for TLS connections, no handshake was done yet so nothing
-         * is written and the connection will just drop. */
-        if (connWrite(conn,err,strlen(err)) == -1) {
-            /* Nothing to do, Just to avoid the warning... */
-        }
-        g_pserver->stat_rejected_conn++;
-        connClose(conn);
-        return;
-    }
-
     /* Limit the number of connections we take at the same time.
      *
      * Admission control will happen before a client is created and connAccept()
@@ -1824,8 +1810,6 @@ int writeToClient(client *c, int handler_installed) {
        is a replica, so only attempt to do so if that's the case. */
     if (c->flags & CLIENT_SLAVE && !(c->flags & CLIENT_MONITOR) && c->replstate == SLAVE_STATE_ONLINE) {
         std::unique_lock<fastlock> repl_backlog_lock (g_pserver->repl_backlog_lock);
-        // Ensure all writes to the repl backlog are visible
-        std::atomic_thread_fence(std::memory_order_acquire);
 
         while (clientHasPendingReplies(c)) {
             long long repl_end_idx = getReplIndexFromOffset(c->repl_end_off);
@@ -1999,9 +1983,9 @@ void ProcessPendingAsyncWrites()
          * writes may have been signalled without having been copied to the replyAsync buffer,
          * thus causing the buffer to be NULL */ 
         if (c->replyAsync != nullptr){
-            size_t size = c->replyAsync->used;
+            int size = c->replyAsync->used;
 
-            if (listLength(c->reply) == 0 && size <= static_cast<size_t>(PROTO_REPLY_CHUNK_BYTES - c->bufpos)) {
+            if (listLength(c->reply) == 0 && size <= (PROTO_REPLY_CHUNK_BYTES - c->bufpos)) {
                 memcpy(c->buf + c->bufpos, c->replyAsync->buf(), size);
                 c->bufpos += size;
             } else {
@@ -2093,6 +2077,8 @@ int handleClientsWithPendingWrites(int iel, int aof_state) {
         * that may trigger write error or recreate handler. */
         if ((flags & CLIENT_PROTECTED) && !(flags & CLIENT_SLAVE)) continue;
 
+        //std::unique_lock<decltype(c->lock)> lock(c->lock);
+
         /* Don't write to clients that are going to be closed anyway. */
         if (c->flags & CLIENT_CLOSE_ASAP) continue;
 
@@ -2110,7 +2096,6 @@ int handleClientsWithPendingWrites(int iel, int aof_state) {
 
         /* If after the synchronous writes above we still have data to
         * output to the client, we need to install the writable handler. */
-        std::unique_lock<decltype(c->lock)> lock(c->lock);
         if (clientHasPendingReplies(c)) {
             if (connSetWriteHandlerWithBarrier(c->conn, sendReplyToClient, ae_flags, true) == C_ERR) {
                 freeClientAsync(c);
@@ -2757,10 +2742,9 @@ void readQueryFromClient(connection *conn) {
         parseClientCommandBuffer(c);
         if (g_pserver->enable_async_commands && !serverTL->disable_async_commands && listLength(g_pserver->monitors) == 0 && (aeLockContention() || serverTL->rgdbSnapshot[c->db->id] || g_fTestMode)) {
             // Frequent writers aren't good candidates for this optimization, they cause us to renew the snapshot too often
-            //  so we exclude them unless the snapshot we need already exists.
-            // Note: In test mode we want to create snapshots as often as possibl to excercise them - we don't care about perf
+            //  so we exclude them unless the snapshot we need already exists
             bool fSnapshotExists = c->db->mvccLastSnapshot >= c->mvccCheckpoint;
-            bool fWriteTooRecent = !g_fTestMode && (((getMvccTstamp() - c->mvccCheckpoint) >> MVCC_MS_SHIFT) < static_cast<uint64_t>(g_pserver->snapshot_slip)/2);
+            bool fWriteTooRecent = (((getMvccTstamp() - c->mvccCheckpoint) >> MVCC_MS_SHIFT) < static_cast<uint64_t>(g_pserver->snapshot_slip)/2);
 
             // The check below avoids running async commands if this is a frequent writer unless a snapshot is already there to service it
             if (!fWriteTooRecent || fSnapshotExists) {

src/object.cpp

@@ -1091,7 +1091,7 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
     mem_total += g_pserver->initial_memory_usage;
 
     mem = 0;
-    if (g_pserver->repl_backlog && g_pserver->repl_backlog != g_pserver->repl_backlog_disk)
+    if (g_pserver->repl_backlog)
         mem += zmalloc_size(g_pserver->repl_backlog);
     mh->repl_backlog = mem;
     mem_total += mem;
@@ -1142,7 +1142,8 @@ struct redisMemOverhead *getMemoryOverheadData(void) {
         mem_total+=mem;
         
         std::unique_lock<fastlock> ul(g_expireLock);
-        mem = db->setexpire()->estimated_bytes_used();
+        mem = db->setexpire()->bytes_used();
+        
         mh->db[mh->num_dbs].overhead_ht_expires = mem;
         mem_total+=mem;
 

src/rdb.cpp

@@ -1657,18 +1657,13 @@ int launchRdbSaveThread(pthread_t &child, rdbSaveInfo *rsi)
 
         g_pserver->rdbThreadVars.tmpfileNum++;
         g_pserver->rdbThreadVars.fRdbThreadCancel = false;
-        pthread_attr_t tattr;
-        pthread_attr_init(&tattr);
-        pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
-        if (pthread_create(&child, &tattr, rdbSaveThread, args)) {
-            pthread_attr_destroy(&tattr);
+        if (pthread_create(&child, NULL, rdbSaveThread, args)) {
             for (int idb = 0; idb < cserver.dbnum; ++idb)
                 g_pserver->db[idb]->endSnapshot(args->rgpdb[idb]);
             args->~rdbSaveThreadArgs();
             zfree(args);
             return C_ERR;
         }
-        pthread_attr_destroy(&tattr);
         g_pserver->child_type = CHILD_TYPE_RDB;
     }
     return C_OK;
@@ -3054,9 +3049,7 @@ void rdbLoadProgressCallback(rio *r, const void *buf, size_t len) {
         (r->keys_since_last_callback >= g_pserver->loading_process_events_interval_keys)))
     {
         rdbAsyncWorkThread *pwthread = reinterpret_cast<rdbAsyncWorkThread*>(r->chksum_arg);
-        mstime_t mstime;
-        __atomic_load(&g_pserver->mstime, &mstime, __ATOMIC_RELAXED);
-        bool fUpdateReplication = (mstime - r->last_update) > 1000;
+        bool fUpdateReplication = (g_pserver->mstime - r->last_update) > 1000;
 
         if (fUpdateReplication) {
             listIter li;
@@ -3115,12 +3108,13 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
     std::unique_ptr<rdbInsertJob> spjob;
 
     // If we're tracking changes we need to reset this
-    std::vector<bool> fTracking(cserver.dbnum);
-    // We don't want to track here because processChangesAsync is outside the normal scope handling
-    for (int idb = 0; idb < cserver.dbnum; ++idb) {
-        if ((fTracking[idb] = g_pserver->db[idb]->FTrackingChanges()))
+    bool fTracking = g_pserver->db[0]->FTrackingChanges();
+    if (fTracking) {
+        // We don't want to track here because processChangesAsync is outside the normal scope handling
+        for (int idb = 0; idb < cserver.dbnum; ++idb) {
             if (g_pserver->db[idb]->processChanges(false))
                 g_pserver->db[idb]->commitChanges();
+        }
     }
 
     rdb->update_cksum = rdbLoadProgressCallback;
@@ -3488,10 +3482,11 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
     }
 
     wqueue.endWork();
-    // Reset track changes
-    for (int idb = 0; idb < cserver.dbnum; ++idb) {
-        if (fTracking[idb])
+    if (fTracking) {
+        // Reset track changes
+        for (int idb = 0; idb < cserver.dbnum; ++idb) {
             g_pserver->db[idb]->trackChanges(false);
+        }
     }
     if (empty_keys_skipped) {
         serverLog(LL_WARNING,
@@ -3509,10 +3504,11 @@ int rdbLoadRio(rio *rdb, int rdbflags, rdbSaveInfo *rsi) {
      * the RDB file from a socket during initial SYNC (diskless replica mode),
      * we'll report the error to the caller, so that we can retry. */
 eoferr:
+    if (fTracking) {
         // Reset track changes
-    for (int idb = 0; idb < cserver.dbnum; ++idb) {
-        if (fTracking[idb])
+        for (int idb = 0; idb < cserver.dbnum; ++idb) {
             g_pserver->db[idb]->trackChanges(false);
+        }
     }
 
     wqueue.endWork();
@@ -3836,11 +3832,7 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
 
     g_pserver->rdbThreadVars.tmpfileNum++;
     g_pserver->rdbThreadVars.fRdbThreadCancel = false;
-    pthread_attr_t tattr;
-    pthread_attr_init(&tattr);
-    pthread_attr_setstacksize(&tattr, 1 << 23); // 8 MB
-    if (pthread_create(&child, &tattr, rdbSaveToSlavesSocketsThread, args)) {
-        pthread_attr_destroy(&tattr);
+    if (pthread_create(&child, nullptr, rdbSaveToSlavesSocketsThread, args)) {
         serverLog(LL_WARNING,"Can't save in background: fork: %s",
             strerror(errno));
 
@@ -3866,7 +3858,6 @@ int rdbSaveToSlavesSockets(rdbSaveInfo *rsi) {
         closeChildInfoPipe();
         return C_ERR;
     }
-    pthread_attr_destroy(&tattr);
     g_pserver->child_type = CHILD_TYPE_RDB;
 
     serverLog(LL_NOTICE,"Background RDB transfer started");

src/redis-cli.c

@@ -5485,8 +5485,8 @@ static int clusterManagerCommandRebalance(int argc, char **argv) {
             listAddNodeTail(lsrc, src);
             table = clusterManagerComputeReshardTable(lsrc, numslots);
             listRelease(lsrc);
-            int table_len = 0;
-            if (!table || (table_len = (int) listLength(table)) != numslots) {
+            int table_len = (int) listLength(table);
+            if (!table || table_len != numslots) {
                 clusterManagerLogErr("*** Assertion failed: Reshard table "
                                      "!= number of slots");
                 result = 0;

src/replication.cpp

@@ -190,7 +190,7 @@ int bg_unlink(const char *filename) {
 bool createDiskBacklog() {
     // Lets create some disk backed pages and add them here
     std::string path = "./repl-backlog-temp" + std::to_string(gettid());
-#if (defined __APPLE__ || defined __FreeBSD__)
+#ifdef __APPLE__
     int fd = open(path.c_str(), O_CREAT | O_RDWR, S_IRUSR | S_IWUSR);
 #else
     int fd = open(path.c_str(), O_CREAT | O_RDWR | O_LARGEFILE, S_IRUSR | S_IWUSR);
@@ -264,8 +264,9 @@ void resizeReplicationBacklog(long long newsize) {
         newsize = CONFIG_REPL_BACKLOG_MIN_SIZE;
     if (g_pserver->repl_backlog_size == newsize) return;
 
+    std::unique_lock<fastlock> repl_backlog_lock (g_pserver->repl_backlog_lock);
+
     if (g_pserver->repl_backlog != NULL) {
-        std::unique_lock<fastlock> repl_backlog_lock(g_pserver->repl_backlog_lock);
         /* What we actually do is to flush the old buffer and realloc a new
          * empty one. It will refill with new data incrementally.
          * The reason is that copying a few gigabytes adds latency and even
@@ -356,7 +357,7 @@ void freeReplicationBacklog(void) {
 void feedReplicationBacklog(const void *ptr, size_t len) {
     serverAssert(GlobalLocksAcquired());
     const unsigned char *p = (const unsigned char*)ptr;
-    std::unique_lock<fastlock> repl_backlog_lock(g_pserver->repl_backlog_lock, std::defer_lock);
+
 
     if (g_pserver->repl_batch_idxStart >= 0) {
         /* We are lower bounded by the lowest replica offset, or the batch offset start if not applicable */
@@ -416,8 +417,6 @@ void feedReplicationBacklog(const void *ptr, size_t len) {
                 // We need to update a few variables or later asserts will notice we dropped data
                 g_pserver->repl_batch_offStart = g_pserver->master_repl_offset + len;
                 g_pserver->repl_lowest_off = -1;
-                if (!repl_backlog_lock.owns_lock())
-                    repl_backlog_lock.lock();   // we need to acquire the lock if we'll be overwriting data that writeToClient may be reading
             }
         }
     }
@@ -1132,14 +1131,6 @@ public:
             replica->repl_put_online_on_ack = 1;
         }
     }
-
-    void abort() {
-        for (auto replica : replicas) {
-            // Close the connection to force a resync
-            freeClientAsync(replica);
-        }
-        replicas.clear();
-    }
 };
 
 int rdbSaveSnapshotForReplication(struct rdbSaveInfo *rsi) {
@@ -1235,11 +1226,7 @@ int rdbSaveSnapshotForReplication(struct rdbSaveInfo *rsi) {
                 retval = C_ERR;
                 break;
             }
-            if (count != snapshotDeclaredCount) {
-                serverLog(LL_WARNING, "Replication BUG: Count of keys sent does not match actual count.  Aborting full sync.");
-                replBuf.abort();
-                break;
-            }
+            serverAssert(count == snapshotDeclaredCount);
         }
         
         replBuf.end();
@@ -2989,11 +2976,10 @@ error:
 }
 
 void readSyncBulkPayload(connection *conn) {
-    serverAssert(GlobalLocksAcquired());
     rdbSaveInfo rsi;
     redisMaster *mi = (redisMaster*)connGetPrivateData(conn);
     static int usemark = 0;
-    if (mi == nullptr || conn != mi->repl_transfer_s) {
+    if (mi == nullptr) {
         // We're about to be free'd so bail out
         return;
     }
@@ -3006,9 +2992,6 @@ void readSyncBulkPayload(connection *conn) {
             return;
     }
 
-    if (conn != mi->repl_transfer_s)
-        return;
-
     /* Final setup of the connected slave <- master link */
     replicationCreateMasterClient(mi,mi->repl_transfer_s,rsi.repl_stream_db);
     if (mi->isRocksdbSnapshotRepl) {
@@ -3817,14 +3800,12 @@ int connectWithMaster(redisMaster *mi) {
  * Never call this function directly, use cancelReplicationHandshake() instead.
  */
 void undoConnectWithMaster(redisMaster *mi) {
-    serverAssert(GlobalLocksAcquired());
     auto conn = mi->repl_transfer_s;
     connSetPrivateData(conn, nullptr);
-    mi->repl_transfer_s = NULL;
-    int result = aePostFunction(g_pserver->rgthreadvar[mi->ielReplTransfer].el, [conn]{
+    aePostFunction(g_pserver->rgthreadvar[mi->ielReplTransfer].el, [conn]{
         connClose(conn);
-    }, false);
-    serverAssert(result == AE_OK);
+    });
+    mi->repl_transfer_s = NULL;
 }
 
 /* Abort the async download of the bulk dataset while SYNC-ing with master.
@@ -3979,10 +3960,6 @@ void freeMasterInfo(redisMaster *mi)
 {
     sdsfree(mi->masterauth);
     zfree(mi->masteruser);
-    if (g_pserver->rdb_filename != nullptr && g_pserver->rdb_filename == mi->repl_transfer_tmpfile) {
-        unlink(g_pserver->rdb_filename);
-        g_pserver->rdb_filename = nullptr;
-    }
     if (mi->repl_transfer_tmpfile)
         zfree(mi->repl_transfer_tmpfile);
     delete mi->staleKeyMap;
@@ -5620,9 +5597,7 @@ void flushReplBacklogToClients()
 
         // Ensure no overflow if we get here
         serverAssert(g_pserver->master_repl_offset - g_pserver->repl_batch_offStart <= g_pserver->repl_backlog_size);
-
-        // Repl backlog writes must become visible to all threads at this point
-        std::atomic_thread_fence(std::memory_order_release);
+        serverAssert(g_pserver->repl_batch_idxStart != g_pserver->repl_backlog_idx);
 
         listIter li;
         listNode *ln;
@@ -5651,10 +5626,7 @@ void flushReplBacklogToClients()
             replica->repl_end_off = g_pserver->master_repl_offset;
 
             /* Only if the there isn't already a pending write do we prepare the client to write */
-            if (replica->repl_curr_off == g_pserver->master_repl_offset) {
-                serverLog(LL_DEBUG, "Pending write when it's on repl_offset=%lld", g_pserver->master_repl_offset);
-                continue;
-            }
+            serverAssert(replica->repl_curr_off != g_pserver->master_repl_offset);
             prepareClientToWrite(replica);
         }
         if (fAsyncWrite)

src/semiorderedset.h

@@ -240,12 +240,14 @@ public:
     bool empty() const noexcept { return celem == 0; }
     size_t size() const noexcept { return celem; }
 
-    size_t estimated_bytes_used() const
+    size_t bytes_used() const
     {
-        // This estimate does't include all the overhead of the internal vectors
-        size_t cb = sizeof(this)
-            + (m_data.capacity() * sizeof(m_data[0]))
-            + sizeof(T) * size();
+        size_t cb = sizeof(this) + (m_data.capacity()-m_data.size())*sizeof(T);
+        for (auto &vec : m_data)
+        {
+            if (vec != nullptr)
+                cb += vec->bytes_used();
+        }
         return cb;
     }
 

src/server.cpp

@@ -779,7 +779,7 @@ struct redisCommand redisCommandTable[] = {
      0,NULL,0,0,0,0,0,0},
 
     {"shutdown",shutdownCommand,-1,
-     "admin no-script ok-loading ok-stale noprop",
+     "admin no-script ok-loading ok-stale",
      0,NULL,0,0,0,0,0,0},
 
     {"lastsave",lastsaveCommand,1,
@@ -2109,10 +2109,8 @@ void databasesCron(bool fMainThread) {
                         aeAcquireLock();
                     }
 
-                    if (serverTL->rehashCtl->done.load(std::memory_order_relaxed)) {
-                        dictCompleteRehashAsync(serverTL->rehashCtl, true /*fFree*/);
-                        serverTL->rehashCtl = nullptr;
-                    }
+                    dictCompleteRehashAsync(serverTL->rehashCtl, true /*fFree*/);
+                    serverTL->rehashCtl = nullptr;
                 }
 
                 serverAssert(serverTL->rehashCtl == nullptr);
@@ -2648,26 +2646,6 @@ int serverCron(struct aeEventLoop *eventLoop, long long id, void *clientData) {
         }
     }
 
-    if (g_pserver->soft_shutdown) {
-        /* Loop through our clients list and see if there are any active clients */
-        listIter li;
-        listNode *ln;
-        listRewind(g_pserver->clients, &li);
-        bool fActiveClient = false;
-        while ((ln = listNext(&li)) && !fActiveClient) {
-            client *c = (client*)listNodeValue(ln);
-            if (c->flags & CLIENT_IGNORE_SOFT_SHUTDOWN)
-                continue;
-            fActiveClient = true;
-        }
-        if (!fActiveClient) {
-            if (prepareForShutdown(SHUTDOWN_NOFLAGS) == C_OK) {
-                serverLog(LL_WARNING, "All active clients have disconnected while a soft shutdown is pending.  Shutting down now.");
-                throw ShutdownException();
-            }
-        }
-    }
-
     g_pserver->cronloops++;
     return 1000/g_pserver->hz;
 }
@@ -2795,8 +2773,6 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
     AeLocker locker;
     int iel = ielFromEventLoop(eventLoop);
 
-    tlsProcessPendingData();
-
     locker.arm();
 
     /* end any snapshots created by fast async commands */
@@ -2827,6 +2803,7 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
         uint64_t processed = 0;
         int aof_state = g_pserver->aof_state;
         locker.disarm();
+        processed += tlsProcessPendingData();
         processed += handleClientsWithPendingWrites(iel, aof_state);
         locker.arm();
         processed += freeClientsInAsyncFreeQueue(iel);
@@ -2837,6 +2814,13 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
     /* Handle precise timeouts of blocked clients. */
     handleBlockedClientsTimeout();
 
+    /* Handle TLS pending data. (must be done before flushAppendOnlyFile) */
+    if (tlsHasPendingData()) {
+        locker.disarm();
+        tlsProcessPendingData();
+        locker.arm();
+    }
+
     /* If tls still has pending unread data don't sleep at all. */
     aeSetDontWait(eventLoop, tlsHasPendingData());
 
@@ -2902,7 +2886,7 @@ void beforeSleep(struct aeEventLoop *eventLoop) {
     static thread_local bool fFirstRun = true;
     // note: we also copy the DB pointer in case a DB swap is done while the lock is released
     std::vector<redisDb*> vecdb;    // note we cache the database pointer in case a dbswap is done while the lock is released
-    if (cserver.storage_memory_model == STORAGE_WRITETHROUGH && !g_pserver->loading)
+    if (cserver.storage_memory_model == STORAGE_WRITETHROUGH && g_pserver->m_pstorageFactory != nullptr && !g_pserver->loading)
     {
         if (!fFirstRun) {
             mstime_t storage_process_latency;
@@ -3015,11 +2999,8 @@ void afterSleep(struct aeEventLoop *eventLoop) {
 
         serverAssert(serverTL->gcEpoch.isReset());
         serverTL->gcEpoch = g_pserver->garbageCollector.startEpoch();
-
-        aeAcquireLock();
         for (int idb = 0; idb < cserver.dbnum; ++idb)
             g_pserver->db[idb]->trackChanges(false);
-        aeReleaseLock();
 
         serverTL->disable_async_commands = false;
     }
@@ -5070,8 +5051,6 @@ bool client::asyncCommand(std::function<void(const redisDbPersistentDataSnapshot
                             std::function<void(const redisDbPersistentDataSnapshot *)> &&postFn) 
 {
     serverAssert(FCorrectThread(this));
-    if (serverTL->in_eval)
-        return false;   // we cannot block clients in EVAL
     const redisDbPersistentDataSnapshot *snapshot = nullptr;
     if (!(this->flags & (CLIENT_MULTI | CLIENT_BLOCKED)))
         snapshot = this->db->createSnapshot(this->mvccCheckpoint, false /* fOptional */);
@@ -5316,11 +5295,6 @@ void pingCommand(client *c) {
         return;
     }
 
-    if (g_pserver->soft_shutdown && !(c->flags & CLIENT_IGNORE_SOFT_SHUTDOWN)) {
-        addReplyError(c, "-SHUTDOWN PENDING");
-        return;
-    }
-
     if (c->flags & CLIENT_PUBSUB && c->resp == 2) {
         addReply(c,shared.mbulkhdr[2]);
         addReplyBulkCBuffer(c,"pong",4);
@@ -5558,8 +5532,6 @@ sds genRedisInfoString(const char *section) {
         }
 
         unsigned int lruclock = g_pserver->lruclock.load();
-        ustime_t ustime;
-        __atomic_load(&g_pserver->ustime, &ustime, __ATOMIC_RELAXED);
         info = sdscatfmt(info,
             "# Server\r\n"
             "redis_version:%s\r\n"
@@ -5602,7 +5574,7 @@ sds genRedisInfoString(const char *section) {
             supervised,
             g_pserver->runid,
             g_pserver->port ? g_pserver->port : g_pserver->tls_port,
-            (int64_t)ustime,
+            (int64_t)g_pserver->ustime,
             (int64_t)uptime,
             (int64_t)(uptime/(3600*24)),
             g_pserver->hz.load(),
@@ -6715,7 +6687,7 @@ static void sigShutdownHandler(int sig) {
      * If we receive the signal the second time, we interpret this as
      * the user really wanting to quit ASAP without waiting to persist
      * on disk. */
-    if ((g_pserver->shutdown_asap || g_pserver->soft_shutdown) && sig == SIGINT) {
+    if (g_pserver->shutdown_asap && sig == SIGINT) {
         serverLogFromHandler(LL_WARNING, "You insist... exiting now.");
         rdbRemoveTempFile(g_pserver->rdbThreadVars.tmpfileNum, 1);
         g_pserver->garbageCollector.shutdown();
@@ -6726,10 +6698,7 @@ static void sigShutdownHandler(int sig) {
     }
 
     serverLogFromHandler(LL_WARNING, msg);
-    if (g_pserver->config_soft_shutdown)
-        g_pserver->soft_shutdown = true;
-    else
-        g_pserver->shutdown_asap = 1;
+    g_pserver->shutdown_asap = 1;
 }
 
 void setupSignalHandlers(void) {
@@ -7635,7 +7604,7 @@ int main(int argc, char **argv) {
         try {
             loadDataFromDisk();
         } catch (ShutdownException) {
-            _Exit(EXIT_SUCCESS);
+            exit(EXIT_SUCCESS);
         }
 
         if (g_pserver->cluster_enabled) {
@@ -7752,9 +7721,7 @@ int main(int argc, char **argv) {
     g_pserver->garbageCollector.shutdown();
     delete g_pserver->m_pstorageFactory;
 
-    // Don't return because we don't want to run any global dtors
-    _Exit(EXIT_SUCCESS);
-    return 0;   // Ensure we're well formed even though this won't get hit
+    return 0;
 }
 
 /* The End */

src/server.h

@@ -508,7 +508,6 @@ extern int configOOMScoreAdjValuesDefaults[CONFIG_OOM_COUNT];
 #define CLIENT_PREVENT_AOF_PROP (1<<19)  /* Don't propagate to AOF. */
 #define CLIENT_PREVENT_REPL_PROP (1<<20)  /* Don't propagate to slaves. */
 #define CLIENT_PREVENT_PROP (CLIENT_PREVENT_AOF_PROP|CLIENT_PREVENT_REPL_PROP)
-#define CLIENT_IGNORE_SOFT_SHUTDOWN (CLIENT_MASTER | CLIENT_SLAVE | CLIENT_BLOCKED | CLIENT_MONITOR)
 #define CLIENT_PENDING_WRITE (1<<21) /* Client has output to send but a write
                                         handler is yet not installed. */
 #define CLIENT_REPLY_OFF (1<<22)   /* Don't send replies to client. */
@@ -2722,9 +2721,6 @@ struct redisServer {
     long long repl_batch_offStart = -1;
     long long repl_batch_idxStart = -1;
 
-    int config_soft_shutdown = false;
-    bool soft_shutdown = false;
-
     /* Lock Contention Ring Buffer */
     static const size_t s_lockContentionSamples = 64;
     uint16_t rglockSamples[s_lockContentionSamples];

src/snapshot.cpp

@@ -26,6 +26,17 @@ public:
     std::vector<dictEntry*> vecde;
 };
 
+void discontinueAsyncRehash(dict *d) {
+    if (d->asyncdata != nullptr) {
+        auto adata = d->asyncdata;
+        while (adata != nullptr) {
+            adata->abondon = true;
+            adata = adata->next;
+        }
+        d->rehashidx = 0;
+    }
+}
+
 const redisDbPersistentDataSnapshot *redisDbPersistentData::createSnapshot(uint64_t mvccCheckpoint, bool fOptional)
 {
     serverAssert(GlobalLocksAcquired());

src/storage/rocksdb.cpp

@@ -1,234 +0,0 @@
-#include "rocksdb.h"
-#include <string>
-#include <sstream>
-#include <mutex>
-#include <unistd.h>
-#include "rocksdbfactor_internal.h"
-
-static const char keyprefix[] = INTERNAL_KEY_PREFIX;
-
-rocksdb::Options DefaultRocksDBOptions();
-extern "C" pid_t gettid();
-
-bool FInternalKey(const char *key, size_t cch)
-{
-    if (cch >= sizeof(INTERNAL_KEY_PREFIX))
-    {
-        if (memcmp(key, keyprefix, sizeof(INTERNAL_KEY_PREFIX)-1) == 0)
-            return true;
-    }
-    return false;
-}
-
-RocksDBStorageProvider::RocksDBStorageProvider(RocksDBStorageFactory *pfactory, std::shared_ptr<rocksdb::DB> &spdb, std::shared_ptr<rocksdb::ColumnFamilyHandle> &spcolfam, const rocksdb::Snapshot *psnapshot, size_t count)
-    : m_pfactory(pfactory), m_spdb(spdb), m_psnapshot(psnapshot), m_spcolfamily(spcolfam), m_count(count)
-{
-    m_readOptionsTemplate = rocksdb::ReadOptions();
-    m_readOptionsTemplate.verify_checksums = false;
-    m_readOptionsTemplate.snapshot = m_psnapshot;
-}
-
-void RocksDBStorageProvider::insert(const char *key, size_t cchKey, void *data, size_t cb, bool fOverwrite)
-{
-    rocksdb::Status status;
-    std::unique_lock<fastlock> l(m_lock);
-    if (m_spbatch != nullptr)
-        status = m_spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(key, cchKey), rocksdb::Slice((const char*)data, cb));
-    else
-        status = m_spdb->Put(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey), rocksdb::Slice((const char*)data, cb));
-    if (!status.ok())
-        throw status.ToString();
-
-    if (!fOverwrite)
-        ++m_count;
-}
-
-void RocksDBStorageProvider::bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem)
-{
-    if (celem >= 16384) {
-        rocksdb::Options options = DefaultRocksDBOptions();
-        rocksdb::SstFileWriter sst_file_writer(rocksdb::EnvOptions(), options, options.comparator);
-        std::string file_path = m_pfactory->getTempFolder() + "/tmpIngest.";
-        file_path += std::to_string(gettid());
-        file_path += ".sst";
-
-        rocksdb::Status s = sst_file_writer.Open(file_path);
-        if (!s.ok())
-            goto LFallback;
-
-        // Insert rows into the SST file, note that inserted keys must be 
-        // strictly increasing (based on options.comparator)
-        for (size_t ielem = 0; ielem < celem; ++ielem) {
-            s = sst_file_writer.Put(rocksdb::Slice(rgkeys[ielem], rgcbkeys[ielem]), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
-            if (!s.ok()) {
-                unlink(file_path.c_str());
-                goto LFallback;
-            }
-        }
-
-        s = sst_file_writer.Finish();
-        if (!s.ok()) {
-            unlink(file_path.c_str());
-            goto LFallback;
-        }
-
-        auto ingestOptions = rocksdb::IngestExternalFileOptions();
-        ingestOptions.move_files = true;
-        ingestOptions.write_global_seqno = false;
-        ingestOptions.failed_move_fall_back_to_copy = false;
-
-        // Ingest the external SST file into the DB
-        s = m_spdb->IngestExternalFile(m_spcolfamily.get(), {file_path}, ingestOptions);
-        if (!s.ok()) {
-            unlink(file_path.c_str());
-            goto LFallback;
-        }
-    } else {
-    LFallback:
-        auto spbatch = std::make_unique<rocksdb::WriteBatch>();
-        for (size_t ielem = 0; ielem < celem; ++ielem) {
-            spbatch->Put(m_spcolfamily.get(), rocksdb::Slice(rgkeys[ielem], rgcbkeys[ielem]), rocksdb::Slice(rgvals[ielem], rgcbvals[ielem]));
-        }
-        m_spdb->Write(WriteOptions(), spbatch.get());
-    }
-
-    std::unique_lock<fastlock> l(m_lock);
-    m_count += celem;
-}
-
-bool RocksDBStorageProvider::erase(const char *key, size_t cchKey)
-{
-    rocksdb::Status status;
-    std::unique_lock<fastlock> l(m_lock);
-    if (m_spbatch != nullptr)
-    {
-        status = m_spbatch->Delete(m_spcolfamily.get(), rocksdb::Slice(key, cchKey));
-    }
-    else
-    {
-        if (!FKeyExists(key, cchKey))
-            return false;
-        status = m_spdb->Delete(WriteOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey));
-    }
-    if (status.ok())
-        --m_count;
-    return status.ok();
-}
-
-void RocksDBStorageProvider::retrieve(const char *key, size_t cchKey, callbackSingle fn) const
-{
-    rocksdb::PinnableSlice slice;
-    auto status = m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cchKey), &slice);
-    if (status.ok())
-        fn(key, cchKey, slice.data(), slice.size());
-}
-
-size_t RocksDBStorageProvider::clear()
-{
-    size_t celem = count();
-    auto status = m_spdb->DropColumnFamily(m_spcolfamily.get());
-    auto strName = m_spcolfamily->GetName();
-
-    rocksdb::ColumnFamilyHandle *handle = nullptr;
-    rocksdb::ColumnFamilyOptions cf_options(m_pfactory->RocksDbOptions());
-    m_spdb->CreateColumnFamily(cf_options, strName, &handle);
-    m_spcolfamily = std::shared_ptr<rocksdb::ColumnFamilyHandle>(handle);
-
-    if (!status.ok())
-        throw status.ToString();
-    m_count = 0;
-    return celem;
-}
-
-size_t RocksDBStorageProvider::count() const
-{
-    std::unique_lock<fastlock> l(m_lock);
-    return m_count;
-}
-
-bool RocksDBStorageProvider::enumerate(callback fn) const
-{
-    std::unique_ptr<rocksdb::Iterator> it = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(ReadOptions(), m_spcolfamily.get()));
-    size_t count = 0;
-    for (it->SeekToFirst(); it->Valid(); it->Next()) {
-        if (FInternalKey(it->key().data(), it->key().size()))
-            continue;
-        ++count;
-        bool fContinue = fn(it->key().data(), it->key().size(), it->value().data(), it->value().size());
-        if (!fContinue)
-            break;
-    }
-    if (!it->Valid() && count != m_count)
-    {
-        if (const_cast<RocksDBStorageProvider*>(this)->m_count != count)
-            printf("WARNING: rocksdb count mismatch");
-        const_cast<RocksDBStorageProvider*>(this)->m_count = count;
-    }
-    assert(it->status().ok()); // Check for any errors found during the scan
-    return !it->Valid();
-}
-
-const IStorage *RocksDBStorageProvider::clone() const
-{
-    std::unique_lock<fastlock> l(m_lock);
-    const rocksdb::Snapshot *psnapshot = const_cast<RocksDBStorageProvider*>(this)->m_spdb->GetSnapshot();
-    return new RocksDBStorageProvider(m_pfactory, const_cast<RocksDBStorageProvider*>(this)->m_spdb, const_cast<RocksDBStorageProvider*>(this)->m_spcolfamily, psnapshot, m_count);
-}
-
-RocksDBStorageProvider::~RocksDBStorageProvider()
-{
-    if (m_spbatch != nullptr)
-        endWriteBatch();
-    
-    if (m_spdb != nullptr && m_psnapshot == nullptr)
-    {
-        insert(count_key, sizeof(count_key), &m_count, sizeof(m_count), false);
-        flush();
-    }
-
-    if (m_spdb != nullptr)
-    {
-        if (m_psnapshot != nullptr)
-            m_spdb->ReleaseSnapshot(m_psnapshot);
-    }
-}
-
-rocksdb::WriteOptions RocksDBStorageProvider::WriteOptions() const
-{
-    auto opt = rocksdb::WriteOptions();
-    return opt;
-}
-
-void RocksDBStorageProvider::beginWriteBatch()
-{
-    m_lock.lock();
-    m_spbatch = std::make_unique<rocksdb::WriteBatch>();
-}
-
-void RocksDBStorageProvider::endWriteBatch()
-{
-    m_spdb->Write(WriteOptions(), m_spbatch.get());
-    m_spbatch = nullptr;
-    m_lock.unlock();
-}
-
-void RocksDBStorageProvider::batch_lock()
-{
-    m_lock.lock();
-}
-
-void RocksDBStorageProvider::batch_unlock()
-{
-    m_lock.unlock();
-}
-
-void RocksDBStorageProvider::flush()
-{
-    m_spdb->SyncWAL();
-}
-
-bool RocksDBStorageProvider::FKeyExists(const char *key, size_t cch) const
-{
-    rocksdb::PinnableSlice slice;
-    return m_spdb->Get(ReadOptions(), m_spcolfamily.get(), rocksdb::Slice(key, cch), &slice).ok();
-}

src/storage/rocksdb.h

@@ -1,56 +0,0 @@
-#pragma once
-
-#include <memory>
-#include "../IStorage.h"
-#include <rocksdb/db.h>
-#include "../fastlock.h"
-
-#define INTERNAL_KEY_PREFIX "\x00\x04\x03\x00\x05\x02\x04"
-static const char count_key[] = INTERNAL_KEY_PREFIX "__keydb__count\1";
-static const char version_key[] = INTERNAL_KEY_PREFIX "__keydb__version\1";
-static const char meta_key[] = INTERNAL_KEY_PREFIX "__keydb__metadata\1";
-class RocksDBStorageFactory;
-
-class RocksDBStorageProvider : public IStorage
-{
-    RocksDBStorageFactory *m_pfactory;
-    std::shared_ptr<rocksdb::DB> m_spdb;    // Note: This must be first so it is deleted last
-    std::unique_ptr<rocksdb::WriteBatch> m_spbatch;
-    const rocksdb::Snapshot *m_psnapshot = nullptr;
-    std::shared_ptr<rocksdb::ColumnFamilyHandle> m_spcolfamily;
-    rocksdb::ReadOptions m_readOptionsTemplate;
-    size_t m_count = 0;
-    mutable fastlock m_lock {"RocksDBStorageProvider"};
-
-public:
-    RocksDBStorageProvider(RocksDBStorageFactory *pfactory, std::shared_ptr<rocksdb::DB> &spdb, std::shared_ptr<rocksdb::ColumnFamilyHandle> &spcolfam, const rocksdb::Snapshot *psnapshot, size_t count);
-    ~RocksDBStorageProvider();
-
-    virtual void insert(const char *key, size_t cchKey, void *data, size_t cb, bool fOverwrite) override;
-    virtual bool erase(const char *key, size_t cchKey) override;
-    virtual void retrieve(const char *key, size_t cchKey, callbackSingle fn) const override;
-    virtual size_t clear() override;
-    virtual bool enumerate(callback fn) const override;
-
-    virtual const IStorage *clone() const override;
-
-    virtual void beginWriteBatch() override;
-    virtual void endWriteBatch() override;
-
-    virtual void bulkInsert(char **rgkeys, size_t *rgcbkeys, char **rgvals, size_t *rgcbvals, size_t celem) override;
-
-    virtual void batch_lock() override;
-    virtual void batch_unlock() override;
-
-    virtual void flush() override;
-
-    size_t count() const override;
-
-protected:
-    bool FKeyExists(const char *key, size_t cchKey) const;
-
-    const rocksdb::ReadOptions &ReadOptions() const { return m_readOptionsTemplate; }
-    rocksdb::WriteOptions WriteOptions() const;
-};
-
-bool FInternalKey(const char *key, size_t cch);

src/storage/rocksdbfactory.cpp

@@ -1,203 +0,0 @@
-#include "rocksdb.h"
-#include "../version.h"
-#include <rocksdb/filter_policy.h>
-#include <rocksdb/table.h>
-#include <rocksdb/utilities/options_util.h>
-#include <rocksdb/sst_file_manager.h>
-#include <rocksdb/utilities/convenience.h>
-#include <rocksdb/slice_transform.h>
-#include "rocksdbfactor_internal.h"
-#include <sys/types.h>
-#include <sys/stat.h> 
-
-rocksdb::Options DefaultRocksDBOptions() {
-    rocksdb::Options options;
-    options.max_background_compactions = 4;
-    options.max_background_flushes = 2;
-    options.bytes_per_sync = 1048576;
-    options.compaction_pri = rocksdb::kMinOverlappingRatio;
-    options.compression = rocksdb::kNoCompression;
-    options.enable_pipelined_write = true;
-    options.allow_mmap_reads = true;
-    options.avoid_unnecessary_blocking_io = true;
-    options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(0));
-
-    rocksdb::BlockBasedTableOptions table_options;
-    table_options.block_size = 16 * 1024;
-    table_options.cache_index_and_filter_blocks = true;
-    table_options.pin_l0_filter_and_index_blocks_in_cache = true;
-    table_options.data_block_index_type = rocksdb::BlockBasedTableOptions::kDataBlockBinaryAndHash;
-    table_options.checksum = rocksdb::kNoChecksum;
-    table_options.format_version = 4;
-    options.table_factory.reset(NewBlockBasedTableFactory(table_options));
-    
-    return options;
-}
-
-IStorageFactory *CreateRocksDBStorageFactory(const char *path, int dbnum, const char *rgchConfig, size_t cchConfig)
-{
-    return new RocksDBStorageFactory(path, dbnum, rgchConfig, cchConfig);
-}
-
-rocksdb::Options RocksDBStorageFactory::RocksDbOptions()
-{
-    rocksdb::Options options = DefaultRocksDBOptions();
-    options.max_open_files = filedsRequired();
-    options.sst_file_manager = m_pfilemanager;
-    options.create_if_missing = true;
-    options.create_missing_column_families = true;
-    options.info_log_level = rocksdb::ERROR_LEVEL;
-    return options;
-}
-
-RocksDBStorageFactory::RocksDBStorageFactory(const char *dbfile, int dbnum, const char *rgchConfig, size_t cchConfig)
-    : m_path(dbfile)
-{
-    dbnum++; // create an extra db for metadata
-    // Get the count of column families in the actual database
-    std::vector<std::string> vecT;
-    auto status = rocksdb::DB::ListColumnFamilies(rocksdb::Options(), dbfile, &vecT);
-    // RocksDB requires we know the count of col families before opening, if the user only wants to see less
-    //  we still have to make room for all column family handles regardless
-    if (status.ok() && (int)vecT.size() > dbnum)
-        dbnum = (int)vecT.size();
-
-    std::vector<rocksdb::ColumnFamilyDescriptor> veccoldesc;
-    veccoldesc.push_back(rocksdb::ColumnFamilyDescriptor(rocksdb::kDefaultColumnFamilyName, rocksdb::ColumnFamilyOptions()));  // ignore default col family
-
-    m_pfilemanager = std::shared_ptr<rocksdb::SstFileManager>(rocksdb::NewSstFileManager(rocksdb::Env::Default()));
-
-    rocksdb::DB *db = nullptr;
-
-    auto options = RocksDbOptions();
-
-    for (int idb = 0; idb < dbnum; ++idb)
-    {
-        rocksdb::ColumnFamilyOptions cf_options(options);
-        cf_options.level_compaction_dynamic_level_bytes = true;
-        veccoldesc.push_back(rocksdb::ColumnFamilyDescriptor(std::to_string(idb), cf_options));
-    }
-
-    if (rgchConfig != nullptr)
-    {
-        std::string options_string(rgchConfig, cchConfig);
-        rocksdb::Status status;
-        if (!(status = rocksdb::GetDBOptionsFromString(options, options_string, &options)).ok())
-        {
-            fprintf(stderr, "Failed to parse FLASH options: %s\r\n", status.ToString().c_str());
-            exit(EXIT_FAILURE);
-        }
-    }
-    
-    std::vector<rocksdb::ColumnFamilyHandle*> handles;
-    status = rocksdb::DB::Open(options, dbfile, veccoldesc, &handles, &db);
-    if (!status.ok())
-        throw status.ToString();
-
-    m_spdb = std::shared_ptr<rocksdb::DB>(db);
-    for (auto handle : handles)
-    {
-        std::string strVersion;
-        auto status = m_spdb->Get(rocksdb::ReadOptions(), handle, rocksdb::Slice(version_key, sizeof(version_key)), &strVersion);
-        if (!status.ok())
-        {
-            setVersion(handle);
-        }
-        else
-        {
-            SymVer ver = parseVersion(strVersion.c_str());
-            auto cmp = compareVersion(&ver);
-            if (cmp == NewerVersion)
-                throw "Cannot load FLASH database created by newer version of KeyDB";
-            if (cmp == OlderVersion)
-                setVersion(handle);
-        }
-        m_vecspcols.emplace_back(handle);
-    }   
-}
-
-RocksDBStorageFactory::~RocksDBStorageFactory()
-{
-    m_spdb->SyncWAL();
-}
-
-void RocksDBStorageFactory::setVersion(rocksdb::ColumnFamilyHandle *handle)
-{
-    auto status = m_spdb->Put(rocksdb::WriteOptions(), handle, rocksdb::Slice(version_key, sizeof(version_key)), rocksdb::Slice(KEYDB_REAL_VERSION, strlen(KEYDB_REAL_VERSION)+1));
-    if (!status.ok())
-        throw status.ToString();
-}
-
-size_t RocksDBStorageFactory::filedsRequired() const {
-    return 256;
-}
-
-std::string RocksDBStorageFactory::getTempFolder()
-{
-    auto path = m_path + "/keydb_tmp/";
-    if (!m_fCreatedTempFolder) {
-        if (!mkdir(path.c_str(), 0700))
-            m_fCreatedTempFolder = true;
-    }
-    return path;
-}
-
-IStorage *RocksDBStorageFactory::createMetadataDb()
-{
-    IStorage *metadataDb = this->create(-1, nullptr, nullptr);
-    metadataDb->insert(meta_key, sizeof(meta_key), (void*)METADATA_DB_IDENTIFIER, strlen(METADATA_DB_IDENTIFIER), false);
-    return metadataDb;
-}
-
-IStorage *RocksDBStorageFactory::create(int db, key_load_iterator iter, void *privdata)
-{
-    ++db;   // skip default col family
-    std::shared_ptr<rocksdb::ColumnFamilyHandle> spcolfamily(m_vecspcols[db].release());
-    size_t count = 0;
-    bool fUnclean = false;
-    
-    std::string value;
-    auto status = m_spdb->Get(rocksdb::ReadOptions(), spcolfamily.get(), rocksdb::Slice(count_key, sizeof(count_key)), &value);
-    if (status.ok() && value.size() == sizeof(size_t))
-    {
-        count = *reinterpret_cast<const size_t*>(value.data());
-        m_spdb->Delete(rocksdb::WriteOptions(), spcolfamily.get(), rocksdb::Slice(count_key, sizeof(count_key)));
-    }
-    else
-    {
-        fUnclean = true;
-    }
-    
-    if (fUnclean || iter != nullptr)
-    {
-        count = 0;
-        auto opts = rocksdb::ReadOptions();
-        opts.tailing = true;
-        std::unique_ptr<rocksdb::Iterator> it = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(opts, spcolfamily.get()));
-
-        it->SeekToFirst();
-        bool fFirstRealKey = true;
-        
-        for (;it->Valid(); it->Next()) {
-            if (FInternalKey(it->key().data(), it->key().size()))
-                continue;
-            if (fUnclean && it->Valid() && fFirstRealKey)
-                printf("\tDatabase %d was not shutdown cleanly, recomputing metrics\n", db);
-            fFirstRealKey = false;
-            if (iter != nullptr)
-                iter(it->key().data(), it->key().size(), privdata);
-            ++count;
-        }
-    }
-    return new RocksDBStorageProvider(this, m_spdb, spcolfamily, nullptr, count);
-}
-
-const char *RocksDBStorageFactory::name() const
-{
-    return "flash";
-}
-
-size_t RocksDBStorageFactory::totalDiskspaceUsed() const
-{
-    return m_pfilemanager->GetTotalSize();
-}

src/storage/teststorageprovider.cpp

@@ -26,10 +26,12 @@ TestStorageProvider::~TestStorageProvider()
 {
 }
 
-void TestStorageProvider::insert(const char *key, size_t cchKey, void *data, size_t cb, bool /*fOverwrite*/)
+void TestStorageProvider::insert(const char *key, size_t cchKey, void *data, size_t cb, bool fOverwrite)
 {
     auto strkey = std::string(key, cchKey);
-    m_map[strkey] = std::string((char*)data, cb);
+    bool fActuallyExists = m_map.find(strkey) != m_map.end();
+    serverAssert(fActuallyExists == fOverwrite);
+    m_map.insert(std::make_pair(strkey, std::string((char*)data, cb)));
 }
 
 

src/t_stream.cpp

@@ -789,13 +789,13 @@ int64_t streamTrim(stream *s, streamAddTrimArgs *args) {
              * update it after (and if) we actually remove the entry */
             unsigned char *pcopy = p;
 
-            int64_t flags = lpGetInteger(p);
+            int flags = lpGetInteger(p);
             p = lpNext(lp, p); /* Skip flags. */
-            int64_t to_skip;
+            int to_skip;
 
-            int64_t ms_delta = lpGetInteger(p);
+            int ms_delta = lpGetInteger(p);
             p = lpNext(lp, p); /* Skip ID ms delta */
-            int64_t seq_delta = lpGetInteger(p);
+            int seq_delta = lpGetInteger(p);
             p = lpNext(lp, p); /* Skip ID seq delta */
 
             streamID currid = {0}; /* For MINID */
@@ -1125,7 +1125,7 @@ int streamIteratorGetID(streamIterator *si, streamID *id, int64_t *numfields) {
              * the first entry emitted for this listpack, then we already
              * emitted the current entry, and have to go back to the previous
              * one. */
-            int64_t lp_count = lpGetInteger(si->lp_ele);
+            int lp_count = lpGetInteger(si->lp_ele);
             while(lp_count--) si->lp_ele = lpPrev(si->lp,si->lp_ele);
             /* Seek lp-count of prev entry. */
             si->lp_ele = lpPrev(si->lp,si->lp_ele);
@@ -1155,7 +1155,7 @@ int streamIteratorGetID(streamIterator *si, streamID *id, int64_t *numfields) {
 
             /* Get the flags entry. */
             si->lp_flags = si->lp_ele;
-            int64_t flags = lpGetInteger(si->lp_ele);
+            int flags = lpGetInteger(si->lp_ele);
             si->lp_ele = lpNext(si->lp,si->lp_ele); /* Seek ID. */
 
             /* Get the ID: it is encoded as difference between the master
@@ -1264,7 +1264,7 @@ void streamIteratorRemoveEntry(streamIterator *si, streamID *current) {
      * deleted entries in the listpack header.
      *
      * We start flagging: */
-    int64_t flags = lpGetInteger(si->lp_flags);
+    int flags = lpGetInteger(si->lp_flags);
     flags |= STREAM_ITEM_FLAG_DELETED;
     lp = lpReplaceInteger(lp,&si->lp_flags,flags);
 

src/tls.cpp

@@ -503,26 +503,11 @@ bool tlsCheckAgainstAllowlist(const char * client){
 #define ASN1_STRING_get0_data ASN1_STRING_data 
 #endif 
 
-class TCleanup {
-    std::function<void()> fn;
-
-public:
-    TCleanup(std::function<void()> fn)
-        : fn(fn)
-    {}
-
-    ~TCleanup() {
-        fn();
-    }
-};
-
 bool tlsValidateCertificateName(tls_connection* conn){
     if (g_pserver->tls_allowlist.empty())
         return true;    // Empty list implies acceptance of all
 
     X509 * cert = SSL_get_peer_certificate(conn->ssl);
-    TCleanup certClen([cert]{X509_free(cert);});
-    
     /* Check the common name (CN) of the certificate first */
     X509_NAME_ENTRY * ne = X509_NAME_get_entry(X509_get_subject_name(cert), X509_NAME_get_index_by_NID(X509_get_subject_name(cert), NID_commonName, -1));
     const char * commonName = reinterpret_cast<const char*>(ASN1_STRING_get0_data(X509_NAME_ENTRY_get_data(ne)));

src/version.h

@@ -1,5 +1,5 @@
-#define KEYDB_REAL_VERSION "255.255.255"
-#define KEYDB_VERSION_NUM 0x00ffffff
+#define KEYDB_REAL_VERSION "6.3.0"
+#define KEYDB_VERSION_NUM 0x00060300
 extern const char *KEYDB_SET_VERSION;   // Unlike real version, this can be overriden by the config
 
 enum VersionCompareResult

tests/integration/psync2-reg-multimaster.tcl

@@ -53,12 +53,12 @@ start_server {overrides {active-replica {yes} multi-master {yes} client-output-b
         set elapsed [expr {[clock milliseconds]-$cycle_start_time}]
         if {$elapsed > $duration*1000} break
         if {rand() < .05} {
-            test "PSYNC2 #3899 regression (multi-master): kill first replica" {
+            test "PSYNC2 #3899 regression: kill first replica" {
                 $R(1) client kill type master
             }
         }
         if {rand() < .05} {
-            test "PSYNC2 #3899 regression (multi-master): kill chained replica" {
+            test "PSYNC2 #3899 regression: kill chained replica" {
                 $R(2) client kill type master
             }
         }
@@ -99,8 +99,4 @@ start_server {overrides {active-replica {yes} multi-master {yes} client-output-b
             fail [format "The three instances have different data sets:\n\tnode 1: %s\n\tnode 2: %s\n\tnode 3: %s\nRun diff -u against /tmp/repldump*.txt for more info" [$R(0) debug digest] [$R(1) debug digest] [$R(2) debug digest] ]
         }
     }
-
-    assert {[s -2 sync_partial_ok] > 0}
-    assert {[s -1 sync_partial_ok] > 0}
-    assert {[s 0 sync_partial_ok] > 0}
 }}}

tests/integration/replication-fast.tcl

@@ -6,7 +6,6 @@ proc prepare_value {size} {
     return $_v
 }
 
-if {$::flash_enabled} {
 start_server {tags {"replication-fast"} overrides {storage-provider {flash ./rocks.db.master} databases 256}} {
     set slave [srv 0 client]
     set slave_host [srv 0 host]
@@ -33,4 +32,3 @@ start_server {tags {"replication-fast"} overrides {storage-provider {flash ./roc
         }
     }
 }
-}

tests/integration/replication-psync-flash.tcl

@@ -8,7 +8,6 @@
 # If reconnect is > 0, the test actually try to break the connection and
 # reconnect with the master, otherwise just the initial synchronization is
 # checked for consistency.
-if {$::flash_enabled} {
 proc test_psync {descr duration backlog_size backlog_ttl delay cond mdl sdl reconnect} {
     start_server [list tags {"repl"} overrides [list storage-provider {flash .rocks.db.m} repl-backlog-size 64m]] {
         start_server [list tags {flash} overrides [list storage-provider {flash ./rocks.db} delete-on-evict no storage-flush-period 10]] {
@@ -134,4 +133,3 @@ foreach mdl {no yes} {
         } $mdl $sdl 1
     }
 }
-}

tests/test_helper.tcl

@@ -47,7 +47,6 @@ set ::all_tests {
     integration/replication-3
     integration/replication-4
     integration/replication-psync
-    integration/replication-psync-flash
     integration/replication-active
     integration/replication-multimaster
     integration/replication-multimaster-connect
@@ -60,13 +59,11 @@ set ::all_tests {
     integration/failover
     integration/keydb-cli
     integration/keydb-benchmark
-    integration/replication-fast
     integration/replication-psync-multimaster
     unit/pubsub
     unit/slowlog
     unit/scripting
     unit/maxmemory
-    unit/flash
     unit/introspection
     unit/introspection-2
     unit/limits
@@ -89,7 +86,6 @@ set ::all_tests {
     integration/logging
     integration/corrupt-dump
     integration/corrupt-dump-fuzzer
-    unit/soft_shutdown
 }
 # Index to the next test to run in the ::all_tests list.
 set ::next_test 0
@@ -739,16 +735,6 @@ for {set j 0} {$j < [llength $argv]} {incr j} {
     }
 }
 
-# Check if we compiled with flash
-set status [catch {exec src/keydb-server --is-flash-enabled}]
-if {$status == 0} {
-    puts "KeyDB was built with FLASH, including FLASH tests"
-    set ::flash_enabled 1
-} else {
-    puts "KeyDB was not built with FLASH.  Excluding FLASH tests"
-    set ::flash_enabled 0
-}
-
 set filtered_tests {}
 
 # Set the filtered tests to be the short list (single_tests) if exists.

tests/unit/flash.tcl

@@ -1,4 +1,3 @@
-if {$::flash_enabled} {
 start_server [list tags {flash} overrides [list storage-provider {flash ./rocks.db} delete-on-evict no storage-flush-period 10]] {
 
     test { FLASH - GET works after eviction } {
@@ -168,4 +167,4 @@ start_server [list tags {flash} overrides [list storage-provider {flash ./rocks.
     }
 
 }
-}
+

tests/unit/memefficiency.tcl

@@ -49,13 +49,10 @@ start_server {tags {"defrag"} overrides {appendonly yes auto-aof-rewrite-percent
             r config set maxmemory 100mb
             r config set maxmemory-policy allkeys-lru
 
-	    for {set i 0} {$i < 10} {incr i} {
-                populate 700000 asdf1 150
-                populate 170000 asdf2 300
-                after 120 ;# serverCron only updates the info once in 100ms
-                set frag [s allocator_frag_ratio]
-		if {$frag >= 1.4} {break}
-	    }
+            populate 700000 asdf1 150
+            populate 170000 asdf2 300
+            after 120 ;# serverCron only updates the info once in 100ms
+            set frag [s allocator_frag_ratio]
             if {$::verbose} {
                 puts "frag $frag"
             }
@@ -510,14 +507,11 @@ start_server {tags {"defrag"} overrides {server-threads 1 active-replica yes} }
             r config set active-defrag-ignore-bytes 2mb
             r config set maxmemory 100mb
             r config set maxmemory-policy allkeys-lru
-            for {set i 0} {$i < 10} {incr i} {
-	        populate 700000 asdf1 150
-                populate 170000 asdf2 300
-                r ping ;# trigger eviction following the previous population
-                after 120 ;# serverCron only updates the info once in 100ms
-                set frag [s allocator_frag_ratio]
-                if {$frag >= 1.4} {break}
-            }
+            populate 700000 asdf1 150
+            populate 170000 asdf2 300
+            r ping ;# trigger eviction following the previous population
+            after 120 ;# serverCron only updates the info once in 100ms
+            set frag [s allocator_frag_ratio]
             if {$::verbose} {
                 puts "frag $frag"
             }

tests/unit/scripting.tcl

@@ -126,10 +126,6 @@ start_server {tags {"scripting"}} {
         r select 9
         set res
     } {original value}
-    
-    test {EVAL background command} {
-	r eval {redis.call("SCAN", "0", "MATCH", "key*", "COUNT", 1000)} 0
-    } {}
 
     if 0 {
         test {EVAL - Script can't run more than configured time limit} {

tests/unit/soft_shutdown.tcl

@@ -1,141 +0,0 @@
-start_server {tags {"soft_shutdown"} } {
-    test {soft shutdown command replies} {
-        assert_equal [r shutdown soft] "OK"
-    }
-
-    test {soft shutdown errors on ping} {
-        catch {[r ping]} e
-        assert_match {SHUTDOWN PENDING} $e
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    test {soft shutdown prevents new connections} {
-        assert_equal [r shutdown soft] "OK"
-        # reconnect
-        set catch_res [catch {set rd [redis_deferring_client]} e]
-        if {$::tls} {
-            assert_equal $catch_res 1
-        } else {
-            assert_match {*SHUTDOWN*} $e
-        }
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    test {soft shutdown allows commands to execute while waiting} {
-        assert_equal [r shutdown soft] "OK"
-        r set test val
-        assert_equal [r get test] {val}
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    test {soft shutdown shuts down after all clients exit} {
-        assert_equal [r shutdown soft] "OK"
-        r close
-        after 500
-        catch {set rd [redis_deferring_client]} e
-        assert_match {*refused*} $e
-    }
-}
-
-start_server {tags {"soft_shutdown"} overrides {soft-shutdown yes} } {
-    test {soft shutdown triggered by SIGINT} {
-        exec kill -SIGINT [s process_id]
-        after 100
-        catch {[r ping]} e
-        assert_match {SHUTDOWN PENDING} $e
-    }
-
-    test {second SIGINT forces a shutdown during a soft shutdown} {
-        exec kill -SIGINT [s process_id]
-        after 100
-        catch {[r ping]} e
-        assert_match {*I/O*} $e
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    test {monitor does not prevent soft shutdown} {
-        set monitor [redis_deferring_client]
-        $monitor monitor
-        assert_equal [r shutdown soft] "OK"
-        r close
-        after 500
-        catch {set rd [redis_deferring_client]} e
-        assert_match {*refused*} $e
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    start_server {} {
-        set node_0 [srv 0 client]
-        set node_0_host [srv 0 host]
-        set node_0_port [srv 0 port]
-        set node_0_pid [srv 0 pid]
-
-        set node_1 [srv -1 client]
-        set node_1_host [srv -1 host]
-        set node_1_port [srv -1 port]
-        set node_1_pid [srv -1 pid]
-
-        $node_0 replicaof $node_1_host $node_1_port
-        wait_for_sync $node_0
-
-        test {soft shutdown works for with master} {
-            $node_1 shutdown soft
-        } {OK}
-
-        test {soft shutdown on master doesn't affect replica} {
-            assert_equal [$node_0 ping] {PONG}
-        }
-
-        test {soft shutdown on master updates ping response} {
-            catch {$node_1 ping} e
-            assert_equal $e {SHUTDOWN PENDING}
-        }
-
-        test {master prevents new connections with soft shutdown} {
-            set c1 [redis $node_1_host $node_1_port 1 $::tls]
-            set catch_res [catch {$c1 read} e]
-            if {$::tls} {
-                assert_equal $catch_res 1
-            } else {
-                assert_match {*SHUTDOWN*} $e
-            }
-        }
-
-        test {master soft shutdown works after all clients disconnect} {
-            $node_1 close
-            after 500
-            catch {set c1 [redis $node_1_host $node_1_port 1 $::tls]} e
-            assert_match {*refused*} $e
-        }
-    }
-}
-
-start_server {tags {"soft_shutdown"} } {
-    start_server {} {
-        set node_0 [srv 0 client]
-        set node_0_host [srv 0 host]
-        set node_0_port [srv 0 port]
-        set node_0_pid [srv 0 pid]
-
-        set node_1 [srv -1 client]
-        set node_1_host [srv -1 host]
-        set node_1_port [srv -1 port]
-        set node_1_pid [srv -1 pid]
-
-        $node_0 replicaof $node_1_host $node_1_port
-        wait_for_sync $node_0
-
-        test {soft shutdown on replica is not blocked by master} {
-            assert_equal [$node_0 shutdown soft] {OK}
-            $node_0 close
-            after 500
-            catch {set c0 [redis $node_0_host $node_0_port 1 $::tls]} e
-            assert_match {*refused*} $e
-        }
-    }
-}

tests/unit/type/stream.tcl

@@ -172,15 +172,6 @@ start_server {
         assert_equal [r XRANGE mystream - +] {{3-0 {f v}} {4-0 {f v}} {5-0 {f v}}}
     }
 
-    test {XTRIM with MINID option, big delta from master record} {
-        r DEL mystream
-        r XADD mystream 1-0 f v
-        r XADD mystream 1641544570597-0 f v
-        r XADD mystream 1641544570597-1 f v
-        r XTRIM mystream MINID 1641544570597-0
-        assert_equal [r XRANGE mystream - +] {{1641544570597-0 {f v}} {1641544570597-1 {f v}}}
-    }
-
     test {XADD mass insertion and XLEN} {
         r DEL mystream
         r multi

utils/redis-copy.rb

@@ -1,27 +1,27 @@
-# redis-copy.rb - Copyright (C) 2009-2010 Salvatore Sanfilippo
+# keydb-copy.rb - Copyright (C) 2009-2010 Salvatore Sanfilippo
 # BSD license, See the COPYING file for more information.
 #
-# Copy the whole dataset from one KeyDB/Redis instance to another one
+# Copy the whole dataset from one Redis instance to another one
 #
 # WARNING: this utility is deprecated and serves as a legacy adapter
-#          for the more-robust redis-copy gem.
+#          for the more-robust keydb-copy gem.
 
 require 'shellwords'
 
 def redisCopy(opts={})
   src = "#{opts[:srchost]}:#{opts[:srcport]}"
   dst = "#{opts[:dsthost]}:#{opts[:dstport]}"
-  `redis-copy #{src.shellescape} #{dst.shellescape}`
+  `keydb-copy #{src.shellescape} #{dst.shellescape}`
 rescue Errno::ENOENT
-  $stderr.puts 'This utility requires the redis-copy executable',
-               'from the redis-copy gem on https://rubygems.org',
-               'To install it, run `gem install redis-copy`.'
+  $stderr.puts 'This utility requires the keydb-copy executable',
+               'from the keydb-copy gem on https://rubygems.org',
+               'To install it, run `gem install keydb-copy`.'
   exit 1
 end
 
-$stderr.puts "This utility is deprecated. Use the redis-copy gem instead."
+$stderr.puts "This utility is deprecated. Use the keydb-copy gem instead."
 if ARGV.length != 4
-    puts "Usage: redis-copy.rb <srchost> <srcport> <dsthost> <dstport>"
+    puts "Usage: keydb-copy.rb <srchost> <srcport> <dsthost> <dstport>"
     exit 1
 end
 puts "WARNING: it's up to you to FLUSHDB the destination host before to continue, press any key when ready."