Redis 2.9.50 (Redis 3.0.0 beta-1)

Logging them at WARNING level was of little utility and of sure disturb.

00-RELEASENOTES

@@ -1,83 +1,45 @@
-Redis 2.6 release notes
+Redis 3.0 release notes
+=======================
 
-Migrating from 2.4 to 2.6
+WARNING: Redis 3.0 is currently a BETA not suitable for production environments.
+
+--------------------------------------------------------------------------------
+Upgrade urgency levels:
+
+LOW:      No need to upgrade unless there are new features you want to use.
+MODERATE: Program an upgrade of the server, but it's not urgent.
+HIGH:     There is a critical bug that may affect a subset of users. Upgrade!
+CRITICAL: There is a critical bug affecting MOST USERS. Upgrade ASAP.
+--------------------------------------------------------------------------------
+
+--[ Redis 3.0.0 Beta 1 (version 2.9.50) ] Release date: 11 Feb 2014
+
+This is the first beta of Redis 3.0.0 (official version is 2.8.50).
+
+The following is a list of improvements in Redis 3.0, compared to Redis 2.8.
+
+* [NEW] Redis Cluster: a distributed implementation of a subset of Redis.
+* [NEW] New "embedded string" object encoding resulting in less cache
+        misses. Big speed gain under certain work loads.
+* [NEW] WAIT command to block waiting for a write to be transmitted to
+        the specified number of slaves.
+* [NEW] MIGRATE connection caching. Much faster keys migraitons.
+* [NEW] MIGARTE new options COPY and REPLACE.
+* [NEW] CLIENT PAUSE command: stop processing client requests for a
+        specified amount of time.
+
+Migrating from 2.8 to 3.0
 =========================
 
-Redis 2.4 is mostly a strict subset of 2.6. However there are a few things
-that you should be aware of:
-
-* You can't use .rdb and AOF files generated with 2.6 into a 2.4 instance.
-* 2.6 slaves can be attached to 2.4 masters, but not the contrary, and only
-  for the time needed to perform the version upgrade.
-
-There are also a few API differences, that are unlikely to cause problems,
-but it is better to keep them in mind:
-
-* SORT now will refuse to sort in numerical mode elements that can't be parsed
-  as numbers.
-* EXPIREs now all have millisecond resolution (but this is very unlikely to
-  break code that was not conceived exploting the previous resolution error
-  in some way.)
-* INFO output is a bit different now, and contains empty lines and comments
-  starting with '#'. All the major clients should be already fixed to work
-  with the new INFO format.
-
-Also the following redis.conf and CONFIG GET / SET parameters changed name:
-
-    * hash-max-zipmap-entries, now replaced by hash-max-ziplist-entries
-    * hash-max-zipmap-value, now replaced by hash-max-ziplist-value
-    * glueoutputbuf was now completely removed as it does not make sense
-
----------
-CHANGELOG
----------
-
-What's new in Redis 2.6.0
-=========================
-
-UPGRADE URGENCY: We suggest new users to start with 2.6.0, and old users to
-                 upgrade after some testing of the application with the new
-                 Redis version.
-
-* Server side Lua scripting, see http://redis.io/commands/eval
-* Virtual Memory removed (was deprecated in 2.4)
-* Hardcoded limits about max number of clients removed.
-* AOF low level semantics is generally more sane, and especially when used
-  in slaves.
-* Milliseconds resolution expires, also added new commands with milliseconds
-  precision (PEXPIRE, PTTL, ...).
-* Clients max output buffer soft and hard limits. You can specifiy different
-  limits for different classes of clients (normal,pubsub,slave).
-* AOF is now able to rewrite aggregate data types using variadic commands,
-  often producing an AOF that is faster to save, load, and is smaller in size.
-* Every redis.conf directive is now accepted as a command line option for the
-  redis-server binary, with the same name and number of arguments.
-* Hash table seed randomization for protection against collisions attacks.
-* Performances improved when writing large objects to Redis.
-* Significant parts of the core refactored or rewritten. New internal APIs
-  and core changes allowed to develop Redis Cluster on top of the new code,
-  however for 2.6 all the cluster code was removed, and will be released with
-  Redis 3.0 when it is more complete and stable.
-* Redis ASCII art logo added at startup.
-* Crash report on memory violation or failed asserts improved significantly
-  to make debugging of hard to catch bugs simpler.
-* redis-benchmark improvements: ability to run selected tests,
-  CSV output, faster, better help.
-* redis-cli improvements: --eval for comfortable development of Lua scripts.
-* SHUTDOWN now supports two optional arguments: "SAVE" and "NOSAVE".
-* INFO output split into sections, the command is now able to just show 
-  pecific sections.
-* New statistics about how many time a command was called, and how much
-  execution time it used (INFO commandstats).
-* More predictable SORT behavior in edge cases.
-* INCRBYFLOAT and HINCRBYFLOAT commands.
+Redis 3.0 is mostly a strict subset of 2.8, you should not have any problem
+upgrading your application from 2.8 to 3.0.
 
 --------------------------------------------------------------------------------
 
-Credits: Where not specified the implementation and design are done by
-Salvatore Sanfilippo and Pieter Noordhuis. Thanks to VMware for making all
-this possible. Also many thanks to all the other contributors and the amazing
-community we have.
+Credits: Where not specified the implementation and design is done by
+Salvatore Sanfilippo. Thanks to Pivotal for making all this possible.
+Also many thanks to all the other contributors and the amazing community
+we have.
 
 See commit messages for more credits.
 

src/cluster.c

@@ -73,21 +73,19 @@ void resetManualFailover(void);
  * Initialization
  * -------------------------------------------------------------------------- */
 
-/* This function is called at startup in order to set the currentEpoch
- * (which is not saved on permanent storage) to the greatest configEpoch found
- * in the loaded nodes (configEpoch is stored on permanent storage as soon as
- * it changes for some node). */
-void clusterSetStartupEpoch() {
+/* Return the greatest configEpoch found in the cluster. */
+uint64_t clusterGetMaxEpoch(void) {
+    uint64_t max = 0;
     dictIterator *di;
     dictEntry *de;
 
     di = dictGetSafeIterator(server.cluster->nodes);
     while((de = dictNext(di)) != NULL) {
         clusterNode *node = dictGetVal(de);
-        if (node->configEpoch > server.cluster->currentEpoch)
-            server.cluster->currentEpoch = node->configEpoch;
+        if (node->configEpoch > max) max = node->configEpoch;
     }
     dictReleaseIterator(di);
+    return max;
 }
 
 int clusterLoadConfig(char *filename) {
@@ -227,7 +225,10 @@ int clusterLoadConfig(char *filename) {
     /* Config sanity check */
     redisAssert(server.cluster->myself != NULL);
     redisLog(REDIS_NOTICE,"Node configuration loaded, I'm %.40s", myself->name);
-    clusterSetStartupEpoch();
+    /* Set the currentEpoch to the max epoch found in the master.
+     * FIXME: this should actually be part of the persistent state, as
+     * documented in the Github issue #1479. */
+    server.cluster->currentEpoch = clusterGetMaxEpoch();
     return REDIS_OK;
 
 fmterr:
@@ -663,7 +664,11 @@ void clusterDelNode(clusterNode *delnode) {
     }
     dictReleaseIterator(di);
 
-    /* 3) Free the node, unlinking it from the cluster. */
+    /* 3) Remove this node from its master's slaves if needed. */
+    if (nodeIsSlave(delnode) && delnode->slaveof)
+        clusterNodeRemoveSlave(delnode->slaveof,delnode);
+
+    /* 4) Free the node, unlinking it from the cluster. */
     freeClusterNode(delnode);
 }
 
@@ -830,10 +835,11 @@ void clearNodeFailureIfNeeded(clusterNode *node) {
 
     /* For slaves we always clear the FAIL flag if we can contact the
      * node again. */
-    if (nodeIsSlave(node)) {
+    if (nodeIsSlave(node) || node->numslots == 0) {
         redisLog(REDIS_NOTICE,
-            "Clear FAIL state for node %.40s: slave is reachable again.",
-                node->name);
+            "Clear FAIL state for node %.40s: %s is reachable again.",
+                node->name,
+                nodeIsSlave(node) ? "slave" : "master without slots");
         node->flags &= ~REDIS_NODE_FAIL;
         clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);
     }
@@ -1111,11 +1117,12 @@ void clusterUpdateSlotsConfigWith(clusterNode *sender, uint64_t senderConfigEpoc
         if (bitmapTestBit(slots,j)) {
             /* We rebind the slot to the new node claiming it if:
              * 1) The slot was unassigned.
-             * 2) The new node claims it with a greater configEpoch. */
-            if (server.cluster->slots[j] == sender) continue;
+             * 2) The new node claims it with a greater configEpoch.
+             * 3) We are not currently importing the slot. */
+            if (server.cluster->slots[j] == sender ||
+                server.cluster->importing_slots_from[j]) continue;
             if (server.cluster->slots[j] == NULL ||
-                server.cluster->slots[j]->configEpoch <
-                senderConfigEpoch)
+                server.cluster->slots[j]->configEpoch < senderConfigEpoch)
             {
                 if (server.cluster->slots[j] == curmaster)
                     newmaster = sender;
@@ -1166,7 +1173,8 @@ int clusterProcessPacket(clusterLink *link) {
         type, (unsigned long) totlen);
 
     /* Perform sanity checks */
-    if (totlen < 8) return 1;
+    if (totlen < 16) return 1; /* At least signature, version, totlen, count. */
+    if (ntohs(hdr->ver) != 0) return 1; /* Can't handle versions other than 0. */
     if (totlen > sdslen(link->rcvbuf)) return 1;
     if (type == CLUSTERMSG_TYPE_PING || type == CLUSTERMSG_TYPE_PONG ||
         type == CLUSTERMSG_TYPE_MEET)
@@ -1360,7 +1368,7 @@ int clusterProcessPacket(clusterLink *link) {
                 }
 
                 /* Master node changed for this slave? */
-                if (sender->slaveof != master) {
+                if (master && sender->slaveof != master) {
                     if (sender->slaveof)
                         clusterNodeRemoveSlave(sender->slaveof,sender);
                     clusterNodeAddSlave(master,sender);
@@ -1426,7 +1434,7 @@ int clusterProcessPacket(clusterLink *link) {
                     if (server.cluster->slots[j]->configEpoch >
                         senderConfigEpoch)
                     {
-                        redisLog(REDIS_WARNING,
+                        redisLog(REDIS_VERBOSE,
                             "Node %.40s has old slots configuration, sending "
                             "an UPDATE message about %.40s",
                                 sender->name, server.cluster->slots[j]->name);
@@ -1579,18 +1587,22 @@ void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
 
     while(1) { /* Read as long as there is data to read. */
         rcvbuflen = sdslen(link->rcvbuf);
-        if (rcvbuflen < 4) {
-            /* First, obtain the first four bytes to get the full message
+        if (rcvbuflen < 8) {
+            /* First, obtain the first 8 bytes to get the full message
              * length. */
-            readlen = 4 - rcvbuflen;
+            readlen = 8 - rcvbuflen;
         } else {
             /* Finally read the full message. */
             hdr = (clusterMsg*) link->rcvbuf;
-            if (rcvbuflen == 4) {
-                /* Perform some sanity check on the message length. */
-                if (ntohl(hdr->totlen) < CLUSTERMSG_MIN_LEN) {
+            if (rcvbuflen == 8) {
+                /* Perform some sanity check on the message signature
+                 * and length. */
+                if (memcmp(hdr->sig,"RCmb",4) != 0 ||
+                    ntohl(hdr->totlen) < CLUSTERMSG_MIN_LEN)
+                {
                     redisLog(REDIS_WARNING,
-                        "Bad message length received from Cluster bus.");
+                        "Bad message length or signature received "
+                        "from Cluster bus.");
                     handleLinkIOError(link);
                     return;
                 }
@@ -1616,7 +1628,7 @@ void clusterReadHandler(aeEventLoop *el, int fd, void *privdata, int mask) {
         }
 
         /* Total length obtained? Process this packet. */
-        if (rcvbuflen >= 4 && rcvbuflen == ntohl(hdr->totlen)) {
+        if (rcvbuflen >= 8 && rcvbuflen == ntohl(hdr->totlen)) {
             if (clusterProcessPacket(link)) {
                 sdsfree(link->rcvbuf);
                 link->rcvbuf = sdsempty();
@@ -1677,6 +1689,10 @@ void clusterBuildMessageHdr(clusterMsg *hdr, int type) {
               myself->slaveof : myself;
 
     memset(hdr,0,sizeof(*hdr));
+    hdr->sig[0] = 'R';
+    hdr->sig[1] = 'C';
+    hdr->sig[2] = 'm';
+    hdr->sig[3] = 'b';
     hdr->type = htons(type);
     memcpy(hdr->sender,myself->name,REDIS_CLUSTER_NAMELEN);
 
@@ -2255,7 +2271,7 @@ void clusterHandleSlaveMigration(int max_slaves) {
         if (nodeIsSlave(node) || nodeFailed(node)) continue;
         okslaves = clusterCountNonFailingSlaves(node);
 
-        if (okslaves == 0 && target == NULL) target = node;
+        if (okslaves == 0 && target == NULL && node->numslots > 0) target = node;
         if (okslaves == max_slaves) {
             for (j = 0; j < node->numslaves; j++) {
                 if (memcmp(node->slaves[j]->name,
@@ -2478,7 +2494,7 @@ void clusterCron(void) {
         if (nodeIsSlave(myself) && nodeIsMaster(node) && !nodeFailed(node)) {
             int okslaves = clusterCountNonFailingSlaves(node);
 
-            if (okslaves == 0) orphaned_masters++;
+            if (okslaves == 0 && node->numslots > 0) orphaned_masters++;
             if (okslaves > max_slaves) max_slaves = okslaves;
             if (nodeIsSlave(myself) && myself->slaveof == node)
                 this_slaves = okslaves;
@@ -3143,10 +3159,10 @@ void clusterCommand(redisClient *c) {
                     return;
                 }
             }
-            /* If this node was the slot owner and the slot was marked as
-             * migrating, assigning the slot to another node will clear
+            /* If this slot is in migrating status but we have no keys
+             * for it assigning the slot to another node will clear
              * the migratig status. */
-            if (server.cluster->slots[slot] == myself &&
+            if (countKeysInSlot(slot) == 0 &&
                 server.cluster->migrating_slots_to[slot])
                 server.cluster->migrating_slots_to[slot] = NULL;
 
@@ -3154,14 +3170,30 @@ void clusterCommand(redisClient *c) {
              * itself also clears the importing status. */
             if (n == myself &&
                 server.cluster->importing_slots_from[slot])
+            {
+                /* This slot was manually migrated, set this node configEpoch
+                 * to a new epoch so that the new version can be propagated
+                 * by the cluster.
+                 *
+                 * FIXME: the new version should be agreed otherwise a race
+                 * is possible if while a manual resharding is in progress
+                 * the master is failed over by a slave. */
+                uint64_t maxEpoch = clusterGetMaxEpoch();
+
+                if (myself->configEpoch != maxEpoch) {
+                    server.cluster->currentEpoch++;
+                    myself->configEpoch = server.cluster->currentEpoch;
+                    clusterDoBeforeSleep(CLUSTER_TODO_FSYNC_CONFIG);
+                }
                 server.cluster->importing_slots_from[slot] = NULL;
+            }
             clusterDelSlot(slot);
             clusterAddSlot(n,slot);
         } else {
             addReplyError(c,"Invalid CLUSTER SETSLOT action or number of arguments");
             return;
         }
-        clusterDoBeforeSleep(CLUSTER_TODO_UPDATE_STATE|CLUSTER_TODO_SAVE_CONFIG);
+        clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG|CLUSTER_TODO_UPDATE_STATE);
         addReply(c,shared.ok);
     } else if (!strcasecmp(c->argv[1]->ptr,"info") && c->argc == 2) {
         /* CLUSTER INFO */

src/cluster.h

@@ -195,7 +195,10 @@ union clusterMsgData {
 
 
 typedef struct {
+    char sig[4];        /* Siganture "RCmb" (Redis Cluster message bus). */
     uint32_t totlen;    /* Total length of this message */
+    uint16_t ver;       /* Protocol version, currently set to 0. */
+    uint16_t notused0;  /* 2 bytes not used. */
     uint16_t type;      /* Message type */
     uint16_t count;     /* Only used for some kind of messages. */
     uint64_t currentEpoch;  /* The epoch accordingly to the sending node. */

src/pubsub.c

@@ -306,8 +306,10 @@ void punsubscribeCommand(redisClient *c) {
 
 void publishCommand(redisClient *c) {
     int receivers = pubsubPublishMessage(c->argv[1],c->argv[2]);
-    if (server.cluster_enabled) clusterPropagatePublish(c->argv[1],c->argv[2]);
-    forceCommandPropagation(c,REDIS_PROPAGATE_REPL);
+    if (server.cluster_enabled)
+        clusterPropagatePublish(c->argv[1],c->argv[2]);
+    else
+        forceCommandPropagation(c,REDIS_PROPAGATE_REPL);
     addReplyLongLong(c,receivers);
 }
 

src/redis-trib.rb

@@ -359,11 +359,11 @@ class RedisTrib
         @nodes.each{|n|
             if n.info[:migrating].size > 0
                 cluster_error \
-                    "[WARNING] Node #{n} has slots in migrating state."
+                    "[WARNING] Node #{n} has slots in migrating state (#{n.info[:migrating].keys.join(",")})."
                 open_slots += n.info[:migrating].keys
             elsif n.info[:importing].size > 0
                 cluster_error \
-                    "[WARNING] Node #{n} has slots in importing state."
+                    "[WARNING] Node #{n} has slots in importing state (#{n.info[:importing].keys.join(",")})."
                 open_slots += n.info[:importing].keys
             end
         }
@@ -469,6 +469,12 @@ class RedisTrib
         #         importing state in 1 slot. That's trivial to address.
         if migrating.length == 1 && importing.length == 1
             move_slot(migrating[0],importing[0],slot,:verbose=>true)
+        elsif migrating.length == 1 && importing.length == 0
+            xputs ">>> Setting #{slot} as STABLE"
+            migrating[0].r.cluster("setslot",slot,"stable")
+        elsif migrating.length == 0 && importing.length == 1
+            xputs ">>> Setting #{slot} as STABLE"
+            importing[0].r.cluster("setslot",slot,"stable")
         else
             xputs "[ERR] Sorry, Redis-trib can't fix this slot yet (work in progress)"
         end
@@ -898,10 +904,10 @@ class RedisTrib
         xputs ">>> Sending CLUSTER FORGET messages to the cluster..."
         @nodes.each{|n|
             next if n == node
-            if n.info[:replicate] && n.info[:replicate].downcase == node_id
+            if n.info[:replicate] && n.info[:replicate].downcase == id
                 # Reconfigure the slave to replicate with some other node
-                xputs ">>> #{n} as replica of #{master}"
                 master = get_master_with_least_replicas
+                xputs ">>> #{n} as replica of #{master}"
                 n.r.cluster("replicate",master.info[:name])
             end
             n.r.cluster("forget",argv[1])

src/version.h

@@ -1 +1 @@
-#define REDIS_VERSION "2.9.11"
+#define REDIS_VERSION "2.9.50"