Fastsync in-mem initial implementation

Author: JohnSully

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, 1, 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/replication.cpp

@@ -1019,8 +1019,6 @@ class replicationBuffer {
     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 @@ class replicationBuffer {
         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 @@ class replicationBuffer {
     }
 
     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_KEYDB_FASTSYNC) && (replicaCur->replstate == SLAVE_STATE_WAIT_BGSAVE_START)) {
-            replica = replicaCur;
-            spreplBuf->addReplica(replica);
+            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_KEYDB_FASTSYNC && 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_KEYDB_FASTSYNC) && 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() &&
@@ -1672,7 +1673,7 @@ 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"))
+            else if (!strcasecmp((const char*)ptrFromObj(c->argv[j+1]), "keydb-fastsync"))
                 c->slave_capa |= SLAVE_CAPA_KEYDB_FASTSYNC;
 
             fCapaCommand = true;
@@ -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_KEYDB_FASTSYNC) && !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;
@@ -3018,13 +3030,11 @@ void readSyncBulkPayload(connection *conn) {
         /* 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;
@@ -3402,7 +3412,7 @@ void parseMasterCapa(redisMaster *mi, sds strcapa)
             // Parse the word
             if (strncmp(szStart, "active-replica", pchEnd - szStart) == 0) {
                 mi->isActive = true;
-            } else if (strncmp(szStart, "rocksdb-snapshot-save", pchEnd - szStart) == 0) {
+            } else if (strncmp(szStart, "keydb-fastsync-save", pchEnd - szStart) == 0) {
                 mi->isKeydbFastsync = true;
             }
             szStart = pchEnd + 1;
@@ -3550,9 +3560,9 @@ void syncWithMaster(connection *conn) {
             "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);
@@ -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);

src/sds.h

@@ -418,7 +418,10 @@ class sdsstring : public sdsview
     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.h

@@ -2688,6 +2688,7 @@ struct redisServer {
 
     int fActiveReplica;                          /* Can this replica also be a master? */
     int fWriteDuringActiveLoad;                  /* Can this active-replica write during an RDB load? */
+    int fEnableFastSync = true;
 
     // Format:
     //  Lower 20 bits: a counter incrementing for each command executed in the same millisecond
@@ -3182,6 +3183,7 @@ 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)
 
@@ -3971,6 +3973,10 @@ 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)

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]