set version to 6.3.4

* fix auditlog test titles, and enable tls tests for CI

* fix tls auditlog key string

build.yaml

@@ -21,13 +21,13 @@ machamp:
         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 4 --verbose 
+        command: ./runtest --clients 4 --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 
+        command: ./runtest-cluster --tls
       sentinel-test:
         type: cmd
         parent: make-build

pkg/deb/master_changelog

@@ -1,3 +1,86 @@
+keydb (6:6.3.4-1distribution_placeholder) codename_placeholder; urgency=medium
+
+  * Add keydb_modstatsd, a module for providing keydb stats to a local statsd server. Can be found under src/modules/keydb_modstatsd
+  * Fixed FreeBSD compile(Thanks to @bra-fsn)
+  * Fixed a bug in s3 bucket config that blocked startup(Thanks to Alexandre Chichmanian)
+  * Fixed a bug causing crash if keys command is called after a blocking command (Fixes issues #659, #605, #532)
+  * Added proper error checking of replica configs, now KeyDB will throw an error if replica-of config is passed before active-replica or multimaster configs
+  * Fixed double free bug in lazy free
+  * Added "overload-protect-percent" config, when enabled this will load shed clients whenever CPU usage exceeds configured value
+  * Added "availability-zone" config, this can be passed any string which will then be reported by info command
+  * Updated eviction logic to account for total system memory availability using sysinfo
+  * Fixed a bug where repl-backlog-size config was modified in keydb.conf with the runtime value during config rewrite
+  * Fixed a possible deadlock when running CLIENT KILL with large number of clients
+  * Fixed a bug where KeyDB would overcommit memory during fork BG save
+  * Fixed a bug with disk repl_backlog causing double free
+  * Fixed integer overflow issue in the temp rdb file naming (Thanks to @karthyuom)
+  * Fixed compile issue with GCC 13.1.1 (Thanks to @michieldwitte)
+  * Removed expireset and restored redis expire behaviour
+  * Fixed a bug causing forked processes to hang, specifically affecting RDB and AOF(Fixes issues #675, #619)
+  * Added "CLUSTER REPLICATE NO ONE" to turn a replica into an empty primary
+  * Added RDB-less full sync, can be enabled with config 'enable-keydb-fastsync'
+  * Added "flash-disable-key-cache" config to disable key cache which stores every key(but not value) in memory
+  * Moved cluster slot to key map to being stored by storage provider, rather than in memory
+  * Moved expires to be stored by storage provider, rather than in memory
+  * Enabled expiry/eviction from storage provider, previously only data stored in memory could be expired/evicted(Fixes #645 along with the 2 previous updates)
+  * Fixed a bug where swapdb result was not recovered after keydb restarts in FLASH mode (Thanks to @karthyuom)
+  * Fixed double free bug when fast sync was canceled early
+  * Fixed a bug where a temp rdb file with zero bytes is generated in flash mode (Thanks to @karthyuom)
+  * Fixed a bug where flash CF options are being reset to default after flushall (Thanks to @karthyuom)(Fixes issue #717)
+  * Updated debug reload command to work with flash
+  * Added support for KEYDB_PASSWORD env variable(Thanks to @einar-pexip)
+  * Compile with flash in Dockerfile(Thanks to @der-eismann)
+  * Fix broken redis-cli symlink (Thanks to @pimvandenbroek)
+  * Use tini for alpine docker(Thanks to @rofafor)
+
+-- Malavan Sotheeswaran <msotheeswaran@snapchat.com>  Sun, 29 Oct 2023 8:39:21 +0000
+
+keydb (6:6.3.3-1distribution_placeholder) codename_placeholder; urgency=medium
+
+  * To help accelerate our development efforts for KeyDB, this will be the last release containing support for Centos 7, Ubuntu 16.04, Ubuntu 18.04, Debian 9 and 32-bit builds
+  * This release contains fixes for 11 issues along with improvements to the KeyDB FLASH feature:
+  * Fixed race condition with expireset access (Issue #597)
+  * Fixed keys command with lua, and added keydb as an alternative to redis as lua variable to access db (Issue #562 and #594)
+  * Fixed hang in aof child (Issue #554)
+  * Fixed leaking fds from RDB save (Issue #453 and #584)
+  * Added config for S3 RDB load/save (Issue #584)
+  * Enabled active defrag during forkless background save to improve average memory efficiency (Issue #460)
+  * Fixed crash with fork background save during replication (Issue #567)
+  * RocksDB has been updated to v7.9.2
+  * Fixed missing slot_to_key map in FLASH cluster mode (Issue #574)
+  * Added keyspace notifications for keys loaded from FLASH at startup
+  * Fixed race condition in prefetchKeysAsync with FLASH enabled (Issue #571)
+  * Fixed integer overflow in rand family of commands(Issue #631, #632, #633)
+  * Fixed bad value in hincrbyfloat(Issue #634)
+  * Fixed OOM hang in rand family of commands(Issue #635)
+  * Added config to limit count of return values in rand family of commands(Issue #636)
+
+-- Ben Schermel <ben@eqalpha.com>  Tue, 18 Apr 2023 10:00:37 +0000
+
+keydb (6:6.3.2-1distribution_placeholder) codename_placeholder; urgency=medium
+
+  * This release contains Beta level support for KeyDB FLASH, new ASYNC commands, latency improvements and a number of bug fixes. 
+  * KeyDB FLASH is included as a Beta feature. FLASH allows KeyDB to persist data to the storage medium it is written to, avoiding the need for AOF/RDB files. KeyDB uses rocksdb as the persistent storage provider and can be enabled with config "storage-provider flash /path/to/rocksdb/output". Read more at https://docs.keydb.dev/docs/flash/
+  * In addition to GET/MGET, ASYNC support has been added for the following commands: HGET, HMGET, HKEYS, HVALS, HGETALL, HSCAN and can be enable with config "enable-async-commands yes”
+  * Packaging support for Ubuntu 22.04 (Jammy) and Debian 12 (Bookworm) has been included with this release
+  * Added new soft shutdown feature, can be enabled with config "soft-shutdown yes".
+  * If soft shutdown is enabled, instead of shutting down right away, the server will wait until all clients have disconnected, and will reject all new connection attempts.
+  * Fixed memory leak with tls certificates when tls allowlist is enabled
+  * Fixed bug in rdb load with flash enabled to ensure all dbs are safe to load (previously only checked db[0])
+  * Fixed race conditions in rdb load and replication
+  * Fixed memory access of rdb file after it should have been deleted
+  * Fixed integer overflow bug in flash(Issue #486)
+  * Improve TLS latency by queueing new commands before executing instead of after
+  * Removed O(n) count of memory usage from info command(replaced by O(1) estimate)
+  * Improved latency of clearing large number of flash DBs(Thanks to Paul Chen for this fix)(Issue #516)
+  * replaced sprintf with snprintf to avoid potential security bugs
+  * Fixed bug where a failed move due to key already existing in move target would result in the key being removed from move source(Thanks to Paul Chen for this fix)(Issue #497)
+  * Fixed usage of deprecated OpenSSL api in OpenSSL v>3.0.1(Issue #392)
+  * Imported security fixes from Redis (CVE-2023-22458 CVE-2022-35977)
+  * Other fixed issues: #480 #477 #454 #452 #303 #425 #492 #541
+
+-- Ben Schermel <ben@eqalpha.com>  Fri, 20 Jan 2023 20:00:37 +0000
+
 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

pkg/docker/Dockerfile_Alpine

@@ -1,4 +1,4 @@
-FROM alpine:3.12
+FROM alpine:3.18
 # 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

src/Makefile

@@ -358,6 +358,7 @@ endif
 OS := $(shell cat /etc/os-release | grep ID= | head -n 1 | cut -d'=' -f2)
 ifeq ($(OS),alpine)
     FINAL_CXXFLAGS+=-DUNW_LOCAL_ONLY
+	FINAL_CXXFLAGS+=-DALPINE
     FINAL_LIBS += -lunwind
 endif
 

src/SnapshotPayloadParseState.cpp

@@ -136,14 +136,33 @@ void SnapshotPayloadParseState::flushQueuedKeys() {
     int idb = current_database;
     serverAssert(vecqueuedKeys.size() == vecqueuedVals.size());
     auto sizePrev = vecqueuedKeys.size();
-    (*insertsInFlight)++;
-    std::weak_ptr<std::atomic<int>> insertsInFlightTmp = insertsInFlight; // C++ GRRRRRRRRRRRRRRRR, we don't want to capute "this" because that's dangerous
     if (current_database < cserver.dbnum) {
-        g_pserver->asyncworkqueue->AddWorkFunction([idb, vecqueuedKeys = std::move(this->vecqueuedKeys), vecqueuedKeysCb = std::move(this->vecqueuedKeysCb), vecqueuedVals = std::move(this->vecqueuedVals), vecqueuedValsCb = std::move(this->vecqueuedValsCb), insertsInFlightTmp, pallocator = m_spallocator.release()]() mutable {
-            g_pserver->db[idb]->bulkDirectStorageInsert(vecqueuedKeys.data(), vecqueuedKeysCb.data(), vecqueuedVals.data(), vecqueuedValsCb.data(), vecqueuedKeys.size());
-            (*(insertsInFlightTmp.lock()))--;
-            delete pallocator;
-        });
+        if (g_pserver->m_pstorageFactory) {
+            (*insertsInFlight)++;
+            std::weak_ptr<std::atomic<int>> insertsInFlightTmp = insertsInFlight; // C++ GRRRRRRRRRRRRRRRR, we don't want to capute "this" because that's dangerous
+            g_pserver->asyncworkqueue->AddWorkFunction([idb, vecqueuedKeys = std::move(this->vecqueuedKeys), vecqueuedKeysCb = std::move(this->vecqueuedKeysCb), vecqueuedVals = std::move(this->vecqueuedVals), vecqueuedValsCb = std::move(this->vecqueuedValsCb), insertsInFlightTmp, pallocator = m_spallocator.release()]() mutable {
+                g_pserver->db[idb]->bulkDirectStorageInsert(vecqueuedKeys.data(), vecqueuedKeysCb.data(), vecqueuedVals.data(), vecqueuedValsCb.data(), vecqueuedKeys.size());
+                (*(insertsInFlightTmp.lock()))--;
+                delete pallocator;
+            });
+        } else {
+            for (size_t ival = 0; ival < vecqueuedKeys.size(); ++ival) {
+                size_t offset = 0;
+                auto spexpire = deserializeExpire(vecqueuedVals[ival], vecqueuedValsCb[ival], &offset);    
+                auto o = deserializeStoredObject(vecqueuedVals[ival] + offset, vecqueuedValsCb[ival] - offset);
+                sds sdsKey = sdsnewlen(vecqueuedKeys[ival], -static_cast<ssize_t>(vecqueuedKeysCb[ival]));
+                if (dbMerge(g_pserver->db[idb], sdsKey, o, false)) {
+                    if (spexpire != nullptr)
+                        g_pserver->db[idb]->setExpire(sdsKey, std::move(*spexpire));
+                } else {
+                    sdsfree(sdsKey);
+                }
+            }
+            vecqueuedKeys.clear();
+            vecqueuedKeysCb.clear();
+            vecqueuedVals.clear();
+            vecqueuedValsCb.clear();
+        }
     } else {
         // else drop the data
         vecqueuedKeys.clear();

src/config.cpp

@@ -2983,6 +2983,7 @@ standardConfig configs[] = {
     createIntConfig("overload-protect-percent", NULL, MODIFIABLE_CONFIG, 0, 200, g_pserver->overload_protect_threshold, 0, INTEGER_CONFIG, NULL, NULL),
     createIntConfig("force-eviction-percent", NULL, MODIFIABLE_CONFIG, 0, 100, g_pserver->force_eviction_percent, 0, INTEGER_CONFIG, NULL, NULL),
     createBoolConfig("enable-async-rehash", NULL, MODIFIABLE_CONFIG, g_pserver->enable_async_rehash, 1, NULL, NULL),
+    createBoolConfig("enable-keydb-fastsync", NULL, MODIFIABLE_CONFIG, g_pserver->fEnableFastSync, 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

@@ -56,7 +56,6 @@ int expireIfNeeded(redisDb *db, robj *key, robj *o);
 void slotToKeyUpdateKeyCore(const char *key, size_t keylen, int add);
 
 std::unique_ptr<expireEntry> deserializeExpire(const char *str, size_t cch, size_t *poffset);
-sds serializeStoredObjectAndExpire(robj_roptr o);
 
 dictType dictChangeDescType {
     dictSdsHash,                /* hash function */

src/networking.cpp

@@ -163,7 +163,6 @@ client *createClient(connection *conn, int iel) {
     c->multibulklen = 0;
     c->bulklen = -1;
     c->sentlen = 0;
-    c->sentlenAsync = 0;
     c->flags = 0;
     c->fPendingAsyncWrite = FALSE;
     c->fPendingAsyncWriteHandler = FALSE;

src/readwritelock.h

@@ -42,9 +42,9 @@ public:
             while (m_readCount > 0)
                 m_cv.wait(rm);
             if (exclusive) {
-                /* Another thread might have the write lock while we have the read lock
-                but won't be able to release it until they can acquire the read lock
-                so release the read lock and try again instead of waiting to avoid deadlock */
+                /* Another thread might have the write lock while we have the internal lock
+                but won't be able to release it until they can acquire the internal lock
+                so release the internal lock and try again instead of waiting to avoid deadlock */
                 while(!m_writeLock.try_lock())
                     m_cv.wait(rm);
             }

src/redis-cli.c

@@ -2524,7 +2524,10 @@ cleanup:
 
 static int clusterManagerNodeConnect(clusterManagerNode *node) {
     if (node->context) redisFree(node->context);
-    node->context = redisConnect(node->ip, node->port);
+    struct timeval tv;
+    tv.tv_sec = config.cluster_manager_command.timeout / 1000;
+    tv.tv_usec = (config.cluster_manager_command.timeout % 1000) * 1000;
+    node->context = redisConnectWithTimeout(node->ip, node->port, tv);
     if (!node->context->err && config.tls) {
         const char *err = NULL;
         if (cliSecureConnection(node->context, config.sslconfig, &err) == REDIS_ERR && err) {
@@ -5626,7 +5629,10 @@ static int clusterManagerCommandImport(int argc, char **argv) {
     char *reply_err = NULL;
     redisReply *src_reply = NULL;
     // Connect to the source node.
-    redisContext *src_ctx = redisConnect(src_ip, src_port);
+    struct timeval tv;
+    tv.tv_sec = config.cluster_manager_command.timeout / 1000;
+    tv.tv_usec = (config.cluster_manager_command.timeout % 1000) * 1000;
+    redisContext *src_ctx = redisConnectWithTimeout(src_ip, src_port, tv);
     if (src_ctx->err) {
         success = 0;
         fprintf(stderr,"Could not connect to KeyDB at %s:%d: %s.\n", src_ip,

src/replication.cpp

@@ -1019,8 +1019,6 @@ public:
     void addReplica(client *replica) {
         replicas.push_back(replica);
         replicationSetupSlaveForFullResync(replica,getPsyncInitialOffset());
-        // Optimize the socket for bulk transfer
-        //connDisableTcpNoDelay(replica->conn);
     }
 
     bool isActive() const { return !replicas.empty(); }
@@ -1093,7 +1091,7 @@ public:
         reply->used += size;
     }
 
-    void addLongWithPrefix(long val, char prefix) {
+    void addLongLongWithPrefix(long long val, char prefix) {
         char buf[128];
         int len;
 
@@ -1105,15 +1103,19 @@ public:
     }
 
     void addArrayLen(int len) {
-        addLongWithPrefix(len, '*');
+        addLongLongWithPrefix(len, '*');
     }
 
     void addLong(long val) {
-        addLongWithPrefix(val, ':');
+        addLongLongWithPrefix(val, ':');
+    }
+
+    void addLongLong(long long val) {
+        addLongLongWithPrefix(val, ':');
     }
 
     void addString(const char *s, unsigned long len) {
-        addLongWithPrefix(len, '$');
+        addLongLongWithPrefix(len, '$');
         addData(s, len);
         addData("\r\n", 2);
     }
@@ -1149,22 +1151,19 @@ int rdbSaveSnapshotForReplication(rdbSaveInfo *rsi) {
     // TODO: This needs to be on a background thread
     int retval = C_OK;
     serverAssert(GlobalLocksAcquired());
-    serverLog(LL_NOTICE, "Starting storage provider fast full sync with target: %s", "disk");
+    serverLog(LL_NOTICE, "Starting fast full sync with target: %s", "disk");
 
     std::shared_ptr<replicationBuffer> spreplBuf = std::make_shared<replicationBuffer>();
     listNode *ln;
     listIter li;
-    client *replica = nullptr;
     listRewind(g_pserver->slaves, &li);
-    while (replica == nullptr && (ln = listNext(&li))) {
+    while ((ln = listNext(&li))) {
         client *replicaCur = (client*)listNodeValue(ln);
-        if ((replicaCur->slave_capa & SLAVE_CAPA_ROCKSDB_SNAPSHOT) && (replicaCur->replstate == SLAVE_STATE_WAIT_BGSAVE_START)) {
-            replica = replicaCur;
-            spreplBuf->addReplica(replica);
+        if ((replicaCur->slave_capa & SLAVE_CAPA_KEYDB_FASTSYNC) && (replicaCur->replstate == SLAVE_STATE_WAIT_BGSAVE_START)) {
+            spreplBuf->addReplica(replicaCur);
             replicaCur->replstate = SLAVE_STATE_FASTSYNC_TX;
         }
     }
-    serverAssert(replica != nullptr);
 
     spreplBuf->addArrayLen(2);   // Two sections: Metadata and databases
 
@@ -1179,7 +1178,7 @@ int rdbSaveSnapshotForReplication(rdbSaveInfo *rsi) {
             spreplBuf->addString(rsi->repl_id, CONFIG_RUN_ID_SIZE);
         spreplBuf->addArrayLen(2);
             spreplBuf->addString("repl-offset", 11);
-            spreplBuf->addLong(rsi->master_repl_offset);
+            spreplBuf->addLongLong(rsi->master_repl_offset);
 
     if (dictSize(g_pserver->lua_scripts)) {
         dictEntry *de;
@@ -1195,13 +1194,13 @@ int rdbSaveSnapshotForReplication(rdbSaveInfo *rsi) {
         di = NULL; /* So that we don't release it again on error. */
     }
 
-    std::shared_ptr<std::vector<std::unique_ptr<const StorageCache>>> spvecspsnapshot = std::make_shared<std::vector<std::unique_ptr<const StorageCache>>>();
+    std::shared_ptr<std::vector<const redisDbPersistentDataSnapshot*>> spvecsnapshot = std::make_shared<std::vector<const redisDbPersistentDataSnapshot*>>();
     for (int idb = 0; idb < cserver.dbnum; ++idb) {
-        spvecspsnapshot->emplace_back(g_pserver->db[idb]->CloneStorageCache());
+        spvecsnapshot->emplace_back(g_pserver->db[idb]->createSnapshot(getMvccTstamp(), false));
     }
     aeReleaseLock();
 
-    g_pserver->asyncworkqueue->AddWorkFunction([spreplBuf = std::move(spreplBuf), spvecspsnapshot = std::move(spvecspsnapshot)]{
+    g_pserver->asyncworkqueue->AddWorkFunction([spreplBuf = std::move(spreplBuf), spvecsnapshot = std::move(spvecsnapshot)]{
         int retval = C_OK;
         auto timeStart = ustime();
         auto lastLogTime = timeStart;
@@ -1211,15 +1210,16 @@ int rdbSaveSnapshotForReplication(rdbSaveInfo *rsi) {
         // Databases
         replBuf.addArrayLen(cserver.dbnum);
         for (int idb = 0; idb < cserver.dbnum; ++idb) {
-            auto &spsnapshot = (*spvecspsnapshot)[idb];
-            size_t snapshotDeclaredCount = spsnapshot->count();
+            auto &spsnapshot = (*spvecsnapshot)[idb];
+            size_t snapshotDeclaredCount = spsnapshot->size();
             replBuf.addArrayLen(snapshotDeclaredCount);
             size_t count = 0;
-            bool result = spsnapshot->enumerate([&replBuf, &count, &cbData, &lastLogTime, &cbLastUpdate](const char *rgchKey, size_t cchKey, const void *rgchVal, size_t cchVal) -> bool{
+            bool result = spsnapshot->iterate_threadsafe([&replBuf, &count, &cbData, &lastLogTime, &cbLastUpdate](const char *strKey, robj_roptr o) -> bool{
                 replBuf.addArrayLen(2);
 
-                replBuf.addString(rgchKey, cchKey);
-                replBuf.addString((const char *)rgchVal, cchVal);
+                replBuf.addString(strKey, sdslen(strKey));
+                sds strT = serializeStoredObjectAndExpire(o);
+                replBuf.addString(strT, sdslen(strT));
                 ++count;
                 if ((count % 8092) == 0) {
                     auto curTime = ustime();
@@ -1230,7 +1230,8 @@ int rdbSaveSnapshotForReplication(rdbSaveInfo *rsi) {
                         lastLogTime = ustime();
                     }
                 }
-                cbData += cchKey + cchVal;
+                cbData += sdslen(strKey) + sdslen(strT);
+                sdsfree(strT);
                 return replBuf.isActive();
             });
 
@@ -1302,7 +1303,7 @@ int startBgsaveForReplication(int mincapa) {
     /* Only do rdbSave* when rsiptr is not NULL,
      * otherwise replica will miss repl-stream-db. */
     if (rsiptr) {
-        if (mincapa & SLAVE_CAPA_ROCKSDB_SNAPSHOT && g_pserver->m_pstorageFactory)
+        if (mincapa & SLAVE_CAPA_KEYDB_FASTSYNC && FFastSyncEnabled())
             retval = rdbSaveSnapshotForReplication(rsiptr);
         else if (socket_target)
             retval = rdbSaveToSlavesSockets(rsiptr);
@@ -1481,7 +1482,7 @@ void syncCommand(client *c) {
     }
 
     /* CASE 0: Fast Sync */
-    if ((c->slave_capa & SLAVE_CAPA_ROCKSDB_SNAPSHOT) && g_pserver->m_pstorageFactory) {
+    if (c->slave_capa & SLAVE_CAPA_KEYDB_FASTSYNC && FFastSyncEnabled()) {
         serverLog(LL_NOTICE,"Fast SYNC on next replication cycle");
     /* CASE 1: BGSAVE is in progress, with disk target. */
     } else if (g_pserver->FRdbSaveInProgress() &&
@@ -1672,8 +1673,8 @@ void replconfCommand(client *c) {
                 c->slave_capa |= SLAVE_CAPA_PSYNC2;
             else if (!strcasecmp((const char*)ptrFromObj(c->argv[j+1]), "activeExpire"))
                 c->slave_capa |= SLAVE_CAPA_ACTIVE_EXPIRE;
-            else if (!strcasecmp((const char*)ptrFromObj(c->argv[j+1]), "rocksdb-snapshot-load"))
-                c->slave_capa |= SLAVE_CAPA_ROCKSDB_SNAPSHOT;
+            else if (!strcasecmp((const char*)ptrFromObj(c->argv[j+1]), "keydb-fastsync"))
+                c->slave_capa |= SLAVE_CAPA_KEYDB_FASTSYNC;
 
             fCapaCommand = true;
         } else if (!strcasecmp((const char*)ptrFromObj(c->argv[j]),"ack")) {
@@ -1738,10 +1739,14 @@ void replconfCommand(client *c) {
 
     if (fCapaCommand) {
         sds reply = sdsnew("+OK");
-        if (g_pserver->fActiveReplica)
+        if (g_pserver->fActiveReplica) {
             reply = sdscat(reply, " active-replica");
-        if (g_pserver->m_pstorageFactory && (c->slave_capa & SLAVE_CAPA_ROCKSDB_SNAPSHOT) && !g_pserver->fActiveReplica)
-            reply = sdscat(reply, " rocksdb-snapshot-save");
+        }
+        if ((c->slave_capa & SLAVE_CAPA_KEYDB_FASTSYNC) && FFastSyncEnabled()) {
+            reply = sdscat(reply, " keydb-fastsync-save");
+        } else {
+            c->slave_capa = (c->slave_capa & (~SLAVE_CAPA_KEYDB_FASTSYNC)); // never try to fast sync for this as they won't expect it
+        }
         reply = sdscat(reply, "\r\n");
         addReplySds(c, reply);
     } else {
@@ -2515,6 +2520,7 @@ bool readSnapshotBulkPayload(connection *conn, redisMaster *mi, rdbSaveInfo &rsi
         }
     }
 
+    serverAssert(mi->parseState != nullptr);
     for (int iter = 0; iter < 10; ++iter) {
         if (mi->parseState->shouldThrottle())
             return false;
@@ -2525,11 +2531,16 @@ bool readSnapshotBulkPayload(connection *conn, redisMaster *mi, rdbSaveInfo &rsi
 
         auto nread = connRead(conn, mi->bulkreadBuffer+qblen, readlen);
         if (nread <= 0) {
-            if (connGetState(conn) != CONN_STATE_CONNECTED)
+            if (connGetState(conn) != CONN_STATE_CONNECTED) {
+                serverLog(LL_WARNING,"I/O error trying to sync with MASTER: %s",
+                    (nread == -1) ? strerror(errno) : "connection lost");
                 cancelReplicationHandshake(mi, true);
+            }
             return false;
         }
+        g_pserver->stat_net_input_bytes += nread;
         mi->repl_transfer_lastio = g_pserver->unixtime;
+        mi->repl_transfer_read += nread;
         sdsIncrLen(mi->bulkreadBuffer,nread);
 
         size_t offset = 0;
@@ -2601,8 +2612,9 @@ bool readSnapshotBulkPayload(connection *conn, redisMaster *mi, rdbSaveInfo &rsi
             rsi.repl_stream_db = mi->parseState->getMetaDataLongLong("repl-stream-db");
             rsi.repl_offset = mi->parseState->getMetaDataLongLong("repl-offset");
             sds str = mi->parseState->getMetaDataStr("repl-id");
-            if (sdslen(str) == CONFIG_RUN_ID_SIZE+1) {
+            if (sdslen(str) == CONFIG_RUN_ID_SIZE) {
                 memcpy(rsi.repl_id, str, CONFIG_RUN_ID_SIZE+1);
+                rsi.repl_id_is_set = 1;
             }
 
             fFinished = true;
@@ -3001,7 +3013,7 @@ void readSyncBulkPayload(connection *conn) {
         return;
     }
 
-    if (mi->isRocksdbSnapshotRepl) {
+    if (mi->isKeydbFastsync) {
         if (!readSnapshotBulkPayload(conn, mi, rsi))
             return;
     } else {
@@ -3014,17 +3026,15 @@ void readSyncBulkPayload(connection *conn) {
 
     /* Final setup of the connected slave <- master link */
     replicationCreateMasterClient(mi,mi->repl_transfer_s,rsi.repl_stream_db);
-    if (mi->isRocksdbSnapshotRepl) {
+    if (mi->isKeydbFastsync) {
         /* We need to handle the case where the initial querybuf data was read by fast sync */
         /* This should match the work readQueryFromClient would do for a master client */
         mi->master->querybuf = sdscatsds(mi->master->querybuf, mi->bulkreadBuffer);
+        mi->master->pending_querybuf = sdscatsds(mi->master->pending_querybuf, mi->bulkreadBuffer);
+        mi->master->read_reploff += sdslen(mi->bulkreadBuffer);
+
         sdsfree(mi->bulkreadBuffer);
         mi->bulkreadBuffer = nullptr;
-
-        mi->master->pending_querybuf = sdscatlen(mi->master->pending_querybuf,
-            mi->master->querybuf,sdslen(mi->master->querybuf));
-
-        mi->master->read_reploff += sdslen(mi->master->querybuf);
     }
     mi->repl_transfer_s = nullptr;
     mi->repl_state = REPL_STATE_CONNECTED;
@@ -3063,13 +3073,13 @@ void readSyncBulkPayload(connection *conn) {
     }
 
     /* Send the initial ACK immediately to put this replica in online state. */
-    if (usemark || mi->isRocksdbSnapshotRepl) replicationSendAck(mi);
+    if (usemark || mi->isKeydbFastsync) replicationSendAck(mi);
 
     /* Restart the AOF subsystem now that we finished the sync. This
      * will trigger an AOF rewrite, and when done will start appending
      * to the new file. */
     if (g_pserver->aof_enabled) restartAOFAfterSYNC();
-    if (mi->isRocksdbSnapshotRepl)
+    if (mi->isKeydbFastsync)
         readQueryFromClient(conn); // There may be querybuf data we just appeneded
     return;
 }
@@ -3395,15 +3405,15 @@ void parseMasterCapa(redisMaster *mi, sds strcapa)
     char *pchEnd = szStart;
 
     mi->isActive = false;
-    mi->isRocksdbSnapshotRepl = false;
+    mi->isKeydbFastsync = false;
     for (;;)
     {
         if (*pchEnd == ' ' || *pchEnd == '\0') {
             // Parse the word
             if (strncmp(szStart, "active-replica", pchEnd - szStart) == 0) {
                 mi->isActive = true;
-            } else if (strncmp(szStart, "rocksdb-snapshot-save", pchEnd - szStart) == 0) {
-                mi->isRocksdbSnapshotRepl = true;
+            } else if (strncmp(szStart, "keydb-fastsync-save", pchEnd - szStart) == 0) {
+                mi->isKeydbFastsync = true;
             }
             szStart = pchEnd + 1;
         }
@@ -3550,9 +3560,9 @@ retry_connect:
             "capa","psync2",
             "capa","activeExpire",
         };
-        if (g_pserver->m_pstorageFactory && !g_pserver->fActiveReplica && g_pserver->repl_diskless_load != REPL_DISKLESS_LOAD_SWAPDB) {
+        if (FFastSyncEnabled() && g_pserver->repl_diskless_load != REPL_DISKLESS_LOAD_SWAPDB) {
             veccapabilities.push_back("capa");
-            veccapabilities.push_back("rocksdb-snapshot-load");
+            veccapabilities.push_back("keydb-fastsync");
         }
 
         err = sendCommandArgv(conn, veccapabilities.size(), veccapabilities.data(), nullptr);
@@ -3735,7 +3745,7 @@ retry_connect:
     }
 
     /* Prepare a suitable temp file for bulk transfer */
-    if (!useDisklessLoad() && !mi->isRocksdbSnapshotRepl) {
+    if (!useDisklessLoad() && !mi->isKeydbFastsync) {
         while(maxtries--) {
             auto dt = std::chrono::system_clock::now().time_since_epoch();
             auto dtMillisecond = std::chrono::duration_cast<std::chrono::milliseconds>(dt);
@@ -3864,6 +3874,10 @@ int cancelReplicationHandshake(redisMaster *mi, int reconnect) {
         delete mi->parseState;
         mi->parseState = nullptr;
     }
+    if (mi->bulkreadBuffer) {
+        sdsfree(mi->bulkreadBuffer);
+        mi->bulkreadBuffer = nullptr;
+    }
 
     if (mi->repl_state == REPL_STATE_TRANSFER) {
         replicationAbortSyncTransfer(mi);
@@ -4329,7 +4343,6 @@ void replicationCacheMaster(redisMaster *mi, client *c) {
     if (c->flags & CLIENT_MULTI) discardTransaction(c);
     listEmpty(c->reply);
     c->sentlen = 0;
-    c->sentlenAsync = 0;
     c->reply_bytes = 0;
     c->bufpos = 0;
     resetClient(c);

src/sds.h

@@ -418,7 +418,10 @@ public:
     sdsstring &operator=(const sdsstring &other)
     {
         sdsfree(m_str);
-        m_str = sdsdup(other.m_str);
+        if (other.m_str != nullptr)
+            m_str = sdsdup(other.m_str);
+        else
+            m_str = nullptr;
         return *this;
     }
 

src/server.cpp

@@ -5102,8 +5102,8 @@ int processCommand(client *c, int callFlags) {
 
             sds str = sdsempty();
             for (int j = 0; j < numkeys; j++) {
-                sdscatsds(str, (sds)ptrFromObj(c->argv[keyindex[j]]));
-                sdscat(str, " ");
+                str = sdscatsds(str, (sds)ptrFromObj(c->argv[keyindex[j]]));
+                str = sdscat(str, " ");
             }
         
             if (numkeys > 0)

src/server.h

@@ -620,7 +620,7 @@ typedef enum {
 #define SLAVE_CAPA_EOF (1<<0)    /* Can parse the RDB EOF streaming format. */
 #define SLAVE_CAPA_PSYNC2 (1<<1) /* Supports PSYNC2 protocol. */
 #define SLAVE_CAPA_ACTIVE_EXPIRE (1<<2) /* Will the slave perform its own expirations? (Don't send delete) */
-#define SLAVE_CAPA_ROCKSDB_SNAPSHOT (1<<3)
+#define SLAVE_CAPA_KEYDB_FASTSYNC (1<<3)
 
 /* Synchronous read timeout - replica side */
 #define CONFIG_REPL_SYNCIO_TIMEOUT 5
@@ -1610,7 +1610,6 @@ struct client {
     unsigned long long reply_bytes; /* Tot bytes of objects in reply list. */
     size_t sentlen;         /* Amount of bytes already sent in the current
                                buffer or object being sent. */
-    size_t sentlenAsync;    /* same as sentlen buf for async buffers (which are a different stream) */
     time_t ctime;           /* Client creation time. */
     long duration;          /* Current command duration. Used for measuring latency of blocking/non-blocking cmds */
     time_t lastinteraction; /* Time of the last interaction, used for timeout */
@@ -1871,7 +1870,7 @@ struct redisMaster {
     long long master_initial_offset;           /* Master PSYNC offset. */
 
     bool isActive = false;
-    bool isRocksdbSnapshotRepl = false;
+    bool isKeydbFastsync = false;
     int repl_state;          /* Replication status if the instance is a replica */
     off_t repl_transfer_size; /* Size of RDB to read from master during sync. */
     off_t repl_transfer_read; /* Amount of RDB read from master during sync. */
@@ -2688,6 +2687,7 @@ struct redisServer {
 
     int fActiveReplica;                          /* Can this replica also be a master? */
     int fWriteDuringActiveLoad;                  /* Can this active-replica write during an RDB load? */
+    int fEnableFastSync = false;
 
     // Format:
     //  Lower 20 bits: a counter incrementing for each command executed in the same millisecond
@@ -3182,6 +3182,7 @@ void trimStringObjectIfNeeded(robj *o);
 robj *deserializeStoredObject(const void *data, size_t cb);
 std::unique_ptr<expireEntry> deserializeExpire(const char *str, size_t cch, size_t *poffset);
 sds serializeStoredObject(robj_roptr o, sds sdsPrefix = nullptr);
+sds serializeStoredObjectAndExpire(robj_roptr o);
 
 #define sdsEncodedObject(objptr) (objptr->encoding == OBJ_ENCODING_RAW || objptr->encoding == OBJ_ENCODING_EMBSTR)
 
@@ -3923,7 +3924,7 @@ void updateActiveReplicaMastersFromRsi(rdbSaveInfo *rsi);
 uint64_t getMvccTstamp();
 void incrementMvccTstamp();
 
-#if __GNUC__ >= 7 && !defined(NO_DEPRECATE_FREE)
+#if __GNUC__ >= 7 && !defined(NO_DEPRECATE_FREE) && !defined(ALPINE)
  [[deprecated]]
 void *calloc(size_t count, size_t size) noexcept;
  [[deprecated]]
@@ -3971,6 +3972,10 @@ inline int ielFromEventLoop(const aeEventLoop *eventLoop)
     return iel;
 }
 
+inline bool FFastSyncEnabled() {
+    return g_pserver->fEnableFastSync && !g_pserver->fActiveReplica;
+}
+
 inline int FCorrectThread(client *c)
 {
     return (c->conn == nullptr)

src/storage/rocksdb.cpp

@@ -7,6 +7,8 @@
 #include "../cluster.h"
 #include "rocksdbfactor_internal.h"
 
+template class std::basic_string<char>;
+
 static const char keyprefix[] = INTERNAL_KEY_PREFIX;
 
 rocksdb::Options DefaultRocksDBOptions();
@@ -24,8 +26,9 @@ bool FInternalKey(const char *key, size_t cch)
 
 std::string getPrefix(unsigned int hashslot)
 {
-    char *hash_char = (char *)&hashslot;
-    return std::string(hash_char + (sizeof(unsigned int) - 2), 2);
+    HASHSLOT_PREFIX_TYPE slot = HASHSLOT_PREFIX_ENDIAN((HASHSLOT_PREFIX_TYPE)hashslot);
+    char *hash_char = (char *)&slot;
+    return std::string(hash_char, HASHSLOT_PREFIX_BYTES);
 }
 
 std::string prefixKey(const char *key, size_t cchKey)
@@ -184,7 +187,7 @@ bool RocksDBStorageProvider::enumerate(callback fn) const
         if (FInternalKey(it->key().data(), it->key().size()))
             continue;
         ++count;
-        bool fContinue = fn(it->key().data()+2, it->key().size()-2, it->value().data(), it->value().size());
+        bool fContinue = fn(it->key().data()+HASHSLOT_PREFIX_BYTES, it->key().size()-HASHSLOT_PREFIX_BYTES, it->value().data(), it->value().size());
         if (!fContinue)
             break;
     }
@@ -203,17 +206,17 @@ bool RocksDBStorageProvider::enumerate_hashslot(callback fn, unsigned int hashsl
     std::string prefix = getPrefix(hashslot);
     std::unique_ptr<rocksdb::Iterator> it = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(ReadOptions(), m_spcolfamily.get()));
     size_t count = 0;
-    for (it->Seek(prefix.c_str()); it->Valid(); it->Next()) {
+    for (it->Seek(prefix); it->Valid(); it->Next()) {
         if (FInternalKey(it->key().data(), it->key().size()))
             continue;
-        if (strncmp(it->key().data(),prefix.c_str(),2) != 0)
+        if (HASHSLOT_PREFIX_RECOVER(*(HASHSLOT_PREFIX_TYPE *)it->key().data()) != hashslot)
             break;
         ++count;
-        bool fContinue = fn(it->key().data()+2, it->key().size()-2, it->value().data(), it->value().size());
+        bool fContinue = fn(it->key().data()+HASHSLOT_PREFIX_BYTES, it->key().size()-HASHSLOT_PREFIX_BYTES, it->value().data(), it->value().size());
         if (!fContinue)
             break;
     }
-    bool full_iter = !it->Valid() || (strncmp(it->key().data(),prefix.c_str(),2) != 0);
+    bool full_iter = !it->Valid() || (HASHSLOT_PREFIX_RECOVER(*(HASHSLOT_PREFIX_TYPE *)it->key().data()) != hashslot);
     if (full_iter && count != g_pserver->cluster->slots_keys_count[hashslot])
     {
         printf("WARNING: rocksdb hashslot count mismatch");
@@ -227,7 +230,8 @@ void RocksDBStorageProvider::setExpire(const char *key, size_t cchKey, long long
 {
     rocksdb::Status status;
     std::unique_lock<fastlock> l(m_lock);
-    std::string prefix((const char *)&expire,sizeof(long long));
+    long long beExpire = htobe64(expire);
+    std::string prefix((const char *)&beExpire,sizeof(long long));
     std::string strKey(key, cchKey);
     if (m_spbatch != nullptr)
         status = m_spbatch->Put(m_spexpirecolfamily.get(), rocksdb::Slice(prefix + strKey), rocksdb::Slice(strKey));
@@ -241,7 +245,8 @@ void RocksDBStorageProvider::removeExpire(const char *key, size_t cchKey, long l
 {
     rocksdb::Status status;
     std::unique_lock<fastlock> l(m_lock);
-    std::string prefix((const char *)&expire,sizeof(long long));
+    long long beExpire = htobe64(expire);
+    std::string prefix((const char *)&beExpire,sizeof(long long));
     std::string strKey(key, cchKey);
     std::string fullKey = prefix + strKey;
     if (!FExpireExists(fullKey))
@@ -278,7 +283,7 @@ std::vector<std::string> RocksDBStorageProvider::getEvictionCandidates(unsigned
         for (it->Seek(randomHashSlot()); it->Valid() && result.size() < count; it->Next()) {
             if (FInternalKey(it->key().data(), it->key().size()))
                 continue;
-            result.emplace_back(it->key().data() + 2, it->key().size() - 2);
+            result.emplace_back(it->key().data() + HASHSLOT_PREFIX_BYTES, it->key().size() - HASHSLOT_PREFIX_BYTES);
         }
     } else {
         std::unique_ptr<rocksdb::Iterator> it = std::unique_ptr<rocksdb::Iterator>(m_spdb->NewIterator(ReadOptions(), m_spexpirecolfamily.get()));

src/storage/rocksdb.h

@@ -4,9 +4,21 @@
 #include "../IStorage.h"
 #include <rocksdb/db.h>
 #include <rocksdb/utilities/write_batch_with_index.h>
+#ifdef __APPLE__
+#include <libkern/OSByteOrder.h>
+#define htole16 OSSwapHostToLittleInt16
+#define le16toh OSSwapLittleToHostInt16
+#define htobe64 OSSwapHostToBigInt64
+#else
+#include <endian.h>
+#endif
 #include "../fastlock.h"
 
 #define INTERNAL_KEY_PREFIX "\x00\x04\x03\x00\x05\x02\x04"
+#define HASHSLOT_PREFIX_TYPE uint16_t
+#define HASHSLOT_PREFIX_BYTES sizeof(HASHSLOT_PREFIX_TYPE)
+#define HASHSLOT_PREFIX_ENDIAN htole16
+#define HASHSLOT_PREFIX_RECOVER le16toh
 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";

src/storage/rocksdbfactory.cpp

@@ -72,7 +72,7 @@ RocksDBStorageFactory::RocksDBStorageFactory(const char *dbfile, int dbnum, cons
     rocksdb::DB *db = nullptr;
 
     auto options = RocksDbOptions();
-    options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(2));
+    options.prefix_extractor.reset(rocksdb::NewFixedPrefixTransform(HASHSLOT_PREFIX_BYTES));
 
     for (int idb = 0; idb < dbnum; ++idb)
     {
@@ -196,7 +196,7 @@ IStorage *RocksDBStorageFactory::create(int db, key_load_iterator iter, void *pr
                 printf("\tDatabase %d was not shutdown cleanly, recomputing metrics\n", db);
             fFirstRealKey = false;
             if (iter != nullptr)
-                iter(it->key().data()+2, it->key().size()-2, privdata);
+                iter(it->key().data()+HASHSLOT_PREFIX_BYTES, it->key().size()-HASHSLOT_PREFIX_BYTES, privdata);
             ++count;
         }
     }

src/version.h

@@ -1,5 +1,5 @@
-#define KEYDB_REAL_VERSION "255.255.255"
-#define KEYDB_VERSION_NUM 0x00ffffff
+#define KEYDB_REAL_VERSION "6.3.4"
+#define KEYDB_VERSION_NUM 0x00060304
 extern const char *KEYDB_SET_VERSION;   // Unlike real version, this can be overriden by the config
 
 enum VersionCompareResult

tests/integration/replication.tcl

@@ -584,6 +584,7 @@ start_server {tags {"repl"}} {
     set master [srv 0 client]
     $master config set repl-diskless-sync yes
     $master config set repl-diskless-sync-delay 1
+    $master config set enable-keydb-fastsync no
     set master_host [srv 0 host]
     set master_port [srv 0 port]
     set master_pid [srv 0 pid]
@@ -727,6 +728,8 @@ start_server {tags {"repl"}} {
             }
         }
     }
+
+    $master config set enable-keydb-fastsync yes
 }
 
 if 0 {
@@ -867,8 +870,10 @@ test {replicaof right after disconnection} {
 test {Kill rdb child process if its dumping RDB is not useful} {
     start_server {tags {"repl"}} {
         set slave1 [srv 0 client]
+        $slave1 config set enable-keydb-fastsync no
         start_server {} {
             set slave2 [srv 0 client]
+            $slave2 config set enable-keydb-fastsync no
             start_server {} {
                 set master [srv 0 client]
                 set master_host [srv 0 host]

tests/unit/maxmemory.tcl

@@ -182,6 +182,8 @@ proc test_slave_buffers {test_name cmd_count payload_len limit_memory pipeline}
                 fail "Replication not started."
             }
 
+            after 100
+
             # measure used memory after the slave connected and set maxmemory
             set orig_used [s -1 used_memory]
             set orig_client_buf [s -1 mem_clients_normal]

tests/unit/tls-auditlog.tcl

@@ -51,7 +51,7 @@ if {$::tls} {
         }
     }
 
-    test {TLS: Able to connect while with SAN having a comprehensive list} {
+    test {TLS Audit Log: Able to connect while with SAN having a comprehensive list} {
         start_server {tags {"tls"} overrides {tls-auditlog-blocklist {dummy.keydb.dev san2.keydb.dev other.keydb.dev}}} {
             catch {r PING} e
             assert_match {PONG} $e
@@ -65,7 +65,7 @@ if {$::tls} {
         }
     }
 
-    test {TLS Audit LogTLS: Able to connect while with SAN having a comprehensive list with wildcards} {
+    test {TLS Audit Log: Able to connect while with SAN having a comprehensive list with wildcards} {
         start_server {tags {"tls"} overrides {tls-auditlog-blocklist {dummy.* san*.dev other.*}}} {
             catch {r PING} e
             assert_match {PONG} $e