Diskless replication: RDB -> slaves transfer draft implementation.

src/rdb.c

@@ -689,6 +689,32 @@ werr:
     return REDIS_ERR;
 }
 
+/* This is just a wrapper to rdbSaveRio() that additionally adds a prefix
+ * and a suffix to the generated RDB dump. The prefix is:
+ *
+ * $EOF:<40 bytes unguessable hex string>\r\n
+ *
+ * While the suffix is the 40 bytes hex string we announced in the prefix.
+ * This way processes receiving the payload can understand when it ends
+ * without doing any processing of the content. */
+int rdbSaveRioWithEOFMark(rio *rdb, int *error) {
+    char eofmark[REDIS_EOF_MARK_SIZE];
+
+    getRandomHexChars(eofmark,REDIS_EOF_MARK_SIZE);
+    if (error) *error = 0;
+    if (rioWrite(rdb,"$EOF:",5) == 0) goto werr;
+    if (rioWrite(rdb,eofmark,REDIS_EOF_MARK_SIZE) == 0) goto werr;
+    if (rioWrite(rdb,"\r\n",2) == 0) goto werr;
+    if (rdbSaveRio(rdb,error) == REDIS_ERR) goto werr;
+    if (rioWrite(rdb,eofmark,REDIS_EOF_MARK_SIZE) == 0) goto werr;
+    return REDIS_OK;
+
+werr: /* Write error. */
+    /* Set 'error' only if not already set by rdbSaveRio() call. */
+    if (error && *error == 0) *error = errno;
+    return REDIS_ERR;
+}
+
 /* Save the DB on disk. Return REDIS_ERR on error, REDIS_OK on success. */
 int rdbSave(char *filename) {
     char tmpfile[256];
@@ -1211,8 +1237,9 @@ eoferr: /* unexpected end of file is handled here with a fatal exit */
     return REDIS_ERR; /* Just to avoid warning */
 }
 
-/* A background saving child (BGSAVE) terminated its work. Handle this. */
-void backgroundSaveDoneHandler(int exitcode, int bysignal) {
+/* A background saving child (BGSAVE) terminated its work. Handle this.
+ * This function covers the case of actual BGSAVEs. */
+void backgroundSaveDoneHandlerDisk(int exitcode, int bysignal) {
     if (!bysignal && exitcode == 0) {
         redisLog(REDIS_NOTICE,
             "Background saving terminated with success");
@@ -1242,7 +1269,113 @@ void backgroundSaveDoneHandler(int exitcode, int bysignal) {
     server.rdb_save_time_start = -1;
     /* Possibly there are slaves waiting for a BGSAVE in order to be served
      * (the first stage of SYNC is a bulk transfer of dump.rdb) */
-    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? REDIS_OK : REDIS_ERR);
+    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? REDIS_OK : REDIS_ERR, REDIS_RDB_CHILD_TYPE_DISK);
+}
+
+/* A background saving child (BGSAVE) terminated its work. Handle this.
+ * This function covers the case of RDB -> Salves socket transfers for
+ * diskless replication. */
+void backgroundSaveDoneHandlerSocket(int exitcode, int bysignal) {
+    if (!bysignal && exitcode == 0) {
+        redisLog(REDIS_NOTICE,
+            "Background RDB transfer terminated with success");
+    } else if (!bysignal && exitcode != 0) {
+        redisLog(REDIS_WARNING, "Background transfer error");
+    } else {
+        redisLog(REDIS_WARNING,
+            "Background transfer terminated by signal %d", bysignal);
+    }
+    server.rdb_child_pid = -1;
+    server.rdb_child_type = REDIS_RDB_CHILD_TYPE_NONE;
+    server.rdb_save_time_start = -1;
+    /* Possibly there are slaves waiting for a BGSAVE in order to be served
+     * (the first stage of SYNC is a bulk transfer of dump.rdb) */
+    updateSlavesWaitingBgsave((!bysignal && exitcode == 0) ? REDIS_OK : REDIS_ERR, REDIS_RDB_CHILD_TYPE_SOCKET);
+}
+
+/* When a background RDB saving/transfer terminates, call the right handler. */
+void backgroundSaveDoneHandler(int exitcode, int bysignal) {
+    switch(server.rdb_child_type) {
+    case REDIS_RDB_CHILD_TYPE_DISK:
+        backgroundSaveDoneHandlerDisk(exitcode,bysignal);
+        break;
+    case REDIS_RDB_CHILD_TYPE_SOCKET:
+        backgroundSaveDoneHandlerSocket(exitcode,bysignal);
+        break;
+    default:
+        redisPanic("Unknown RDB child type.");
+        break;
+    }
+}
+
+/* Spawn an RDB child that writes the RDB to the sockets of the slaves
+ * that are currently in REDIS_REPL_WAIT_BGSAVE_START state. */
+int rdbSaveToSlavesSockets(void) {
+    int *fds;
+    int numfds;
+    listNode *ln;
+    listIter li;
+    pid_t childpid;
+    long long start;
+
+    if (server.rdb_child_pid != -1) return REDIS_ERR;
+
+    fds = zmalloc(sizeof(int)*listLength(server.slaves));
+    numfds = 0;
+
+    listRewind(server.slaves,&li);
+    while((ln = listNext(&li))) {
+        redisClient *slave = ln->value;
+
+        if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) {
+            fds[numfds++] = slave->fd;
+            slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
+        }
+    }
+
+    /* Fork ... */
+    start = ustime();
+    if ((childpid = fork()) == 0) {
+        /* Child */
+        int retval;
+        rio slave_sockets;
+
+        rioInitWithFdset(&slave_sockets,fds,numfds);
+        zfree(fds);
+
+        closeListeningSockets(0);
+        redisSetProcTitle("redis-rdb-to-slaves");
+
+        retval = rdbSaveRioWithEOFMark(&slave_sockets,NULL);
+        if (retval == REDIS_OK) {
+            size_t private_dirty = zmalloc_get_private_dirty();
+
+            if (private_dirty) {
+                redisLog(REDIS_NOTICE,
+                    "RDB: %zu MB of memory used by copy-on-write",
+                    private_dirty/(1024*1024));
+            }
+        }
+        exitFromChild((retval == REDIS_OK) ? 0 : 1);
+    } else {
+        /* Parent */
+        server.stat_fork_time = ustime()-start;
+        server.stat_fork_rate = (double) zmalloc_used_memory() * 1000000 / server.stat_fork_time / (1024*1024*1024); /* GB per second. */
+        latencyAddSampleIfNeeded("fork",server.stat_fork_time/1000);
+        if (childpid == -1) {
+            redisLog(REDIS_WARNING,"Can't save in background: fork: %s",
+                strerror(errno));
+            return REDIS_ERR;
+        }
+        redisLog(REDIS_NOTICE,"Background RDB transfer started by pid %d",childpid);
+        server.rdb_save_time_start = time(NULL);
+        server.rdb_child_pid = childpid;
+        server.rdb_child_type = REDIS_RDB_CHILD_TYPE_SOCKET;
+        updateDictResizePolicy();
+        zfree(fds);
+        return REDIS_OK;
+    }
+    return REDIS_OK; /* unreached */
 }
 
 void saveCommand(redisClient *c) {

src/rdb.h

@@ -101,6 +101,7 @@ int rdbSaveObjectType(rio *rdb, robj *o);
 int rdbLoadObjectType(rio *rdb);
 int rdbLoad(char *filename);
 int rdbSaveBackground(char *filename);
+int rdbSaveToSlavesSockets(void);
 void rdbRemoveTempFile(pid_t childpid);
 int rdbSave(char *filename);
 int rdbSaveObject(rio *rdb, robj *o);

src/redis.c

@@ -1480,6 +1480,7 @@ void initServerConfig(void) {
     server.repl_slave_ro = REDIS_DEFAULT_SLAVE_READ_ONLY;
     server.repl_down_since = 0; /* Never connected, repl is down since EVER. */
     server.repl_disable_tcp_nodelay = REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY;
+    server.repl_diskless = REDIS_DEFAULT_RDB_DISKLESS;
     server.slave_priority = REDIS_DEFAULT_SLAVE_PRIORITY;
     server.master_repl_offset = 0;
 

src/redis.h

@@ -96,6 +96,7 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_REPL_TIMEOUT 60
 #define REDIS_REPL_PING_SLAVE_PERIOD 10
 #define REDIS_RUN_ID_SIZE 40
+#define REDIS_EOF_MARK_SIZE 40
 #define REDIS_OPS_SEC_SAMPLES 16
 #define REDIS_DEFAULT_REPL_BACKLOG_SIZE (1024*1024)    /* 1mb */
 #define REDIS_DEFAULT_REPL_BACKLOG_TIME_LIMIT (60*60)  /* 1 hour */
@@ -113,6 +114,8 @@ typedef long long mstime_t; /* millisecond time type. */
 #define REDIS_DEFAULT_RDB_COMPRESSION 1
 #define REDIS_DEFAULT_RDB_CHECKSUM 1
 #define REDIS_DEFAULT_RDB_FILENAME "dump.rdb"
+#define REDIS_DEFAULT_RDB_DISKLESS 0
+#define REIDS_DEFAULT_RDB_DISKLESS_DELAY 5
 #define REDIS_DEFAULT_SLAVE_SERVE_STALE_DATA 1
 #define REDIS_DEFAULT_SLAVE_READ_ONLY 1
 #define REDIS_DEFAULT_REPL_DISABLE_TCP_NODELAY 0
@@ -796,6 +799,7 @@ struct redisServer {
     int repl_min_slaves_to_write;   /* Min number of slaves to write. */
     int repl_min_slaves_max_lag;    /* Max lag of <count> slaves to write. */
     int repl_good_slaves_count;     /* Number of slaves with lag <= max_lag. */
+    int repl_diskless;              /* Send RDB to slaves sockets directly. */
     /* Replication (slave) */
     char *masterauth;               /* AUTH with this password with master */
     char *masterhost;               /* Hostname of master */
@@ -1138,7 +1142,7 @@ ssize_t syncReadLine(int fd, char *ptr, ssize_t size, long long timeout);
 /* Replication */
 void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc);
 void replicationFeedMonitors(redisClient *c, list *monitors, int dictid, robj **argv, int argc);
-void updateSlavesWaitingBgsave(int bgsaveerr);
+void updateSlavesWaitingBgsave(int bgsaveerr, int type);
 void replicationCron(void);
 void replicationHandleMasterDisconnection(void);
 void replicationCacheMaster(redisClient *c);

src/replication.c

@@ -408,6 +408,28 @@ need_full_resync:
     return REDIS_ERR;
 }
 
+/* Start a BGSAVE for replication goals, which is, selecting the disk or
+ * socket target depending on the configuration, and making sure that
+ * the script cache is flushed before to start.
+ *
+ * Returns REDIS_OK on success or REDIS_ERR otherwise. */
+int startBgsaveForReplication(void) {
+    int retval;
+
+    redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC with target: %s",
+        server.repl_diskless ? "slaves sockets" : "disk");
+
+    if (server.repl_diskless)
+        retval = rdbSaveToSlavesSockets();
+    else
+        retval = rdbSaveBackground(server.rdb_filename);
+
+    /* Flush the script cache, since we need that slave differences are
+     * accumulated without requiring slaves to match our cached scripts. */
+    if (retval == REDIS_OK) replicationScriptCacheFlush();
+    return retval;
+}
+
 /* SYNC and PSYNC command implemenation. */
 void syncCommand(redisClient *c) {
     /* ignore SYNC if already slave or in monitor mode */
@@ -465,7 +487,9 @@ void syncCommand(redisClient *c) {
 
     /* Here we need to check if there is a background saving operation
      * in progress, or if it is required to start one */
-    if (server.rdb_child_pid != -1) {
+    if (server.rdb_child_pid != -1 &&
+        server.rdb_child_type == REDIS_RDB_CHILD_TYPE_DISK)
+    {
         /* Ok a background save is in progress. Let's check if it is a good
          * one for replication, i.e. if there is another slave that is
          * registering differences since the server forked to save. */
@@ -480,12 +504,7 @@ void syncCommand(redisClient *c) {
         }
         if (ln) {
             /* Perfect, the server is already registering differences for
-             * another slave. Set the right state, and copy the buffer.
-             *
-             * Note that if we found a slave in WAIT_BGSAVE_END state, this
-             * means that the current child is of type
-             * REDIS_RDB_CHILD_TYPE_DISK, since the first slave in this state
-             * can only be added when an RDB save with disk target is started. */
+             * another slave. Set the right state, and copy the buffer. */
             copyClientOutputBuffer(c,slave);
             c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
             redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC");
@@ -495,17 +514,31 @@ void syncCommand(redisClient *c) {
             c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
             redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
         }
+    } else if (server.rdb_child_pid != -1 &&
+               server.rdb_child_type == REDIS_RDB_CHILD_TYPE_SOCKET)
+    {
+        /* There is an RDB child process but it is writing directly to
+         * children sockets. We need to wait for the next BGSAVE
+         * in order to synchronize. */
+        c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
+        redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
     } else {
-        /* Ok we don't have a BGSAVE in progress, let's start one. */
-        redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC");
-        if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
-            redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
-            addReplyError(c,"Unable to perform background save");
-            return;
+        if (server.repl_diskless) {
+            /* Diskless replication RDB child is created inside
+             * replicationCron() since we want to delay its start a
+             * few seconds to wait for more slaves to arrive. */
+            c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
+            redisLog(REDIS_NOTICE,"Delay next BGSAVE for SYNC");
+        } else {
+            /* Ok we don't have a BGSAVE in progress, let's start one. */
+            redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC");
+            if (startBgsaveForReplication() != REDIS_OK) {
+                redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
+                addReplyError(c,"Unable to perform background save");
+                return;
+            }
+            c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
         }
-        c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
-        /* Flush the script cache for the new slave. */
-        replicationScriptCacheFlush();
     }
 
     if (server.repl_disable_tcp_nodelay)
@@ -644,10 +677,15 @@ void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) {
 /* This function is called at the end of every background saving.
  * The argument bgsaveerr is REDIS_OK if the background saving succeeded
  * otherwise REDIS_ERR is passed to the function.
+ * The 'type' argument is the type of the child that terminated
+ * (if it had a disk or socket target).
  *
  * The goal of this function is to handle slaves waiting for a successful
- * background saving in order to perform non-blocking synchronization. */
-void updateSlavesWaitingBgsave(int bgsaveerr) {
+ * background saving in order to perform non-blocking synchronization, and
+ * to schedule a new BGSAVE if there are slaves that attached while a
+ * BGSAVE was in progress, but it was not a good one for replication (no
+ * other slave was accumulating differences). */
+void updateSlavesWaitingBgsave(int bgsaveerr, int type) {
     listNode *ln;
     int startbgsave = 0;
     listIter li;
@@ -687,12 +725,7 @@ void updateSlavesWaitingBgsave(int bgsaveerr) {
         }
     }
     if (startbgsave) {
-        /* Since we are starting a new background save for one or more slaves,
-         * we flush the Replication Script Cache to use EVAL to propagate every
-         * new EVALSHA for the first time, since all the new slaves don't know
-         * about previous scripts. */
-        replicationScriptCacheFlush();
-        if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
+        if (startBgsaveForReplication() != REDIS_OK) {
             listIter li;
 
             listRewind(server.slaves,&li);