From 506f9a42b07649d793b9c3245da14aa0fecd2d59 Mon Sep 17 00:00:00 2001
From: antirez <antirez@gmail.com>
Date: Wed, 20 Mar 2013 10:48:42 +0100
Subject: [PATCH] Cluster: new flag PROMOTED introduced.

A slave node set this flag for itself when, after receiving authorization
from the majority of nodes, it turns itself into a master.

At the same time now this flag is tested by nodes receiving a PING
message before reconfiguring after a failover event. This makes the
system more robust: even if currently there is no way to manually turn
a slave into a master it is possible that we'll have such a feature in
the future, or that simply because of misconfiguration a node joins the
cluster as master while others believe it's a slave. This alone is now
no longer enough to trigger reconfiguration as other nodes will check
for the PROMOTED flag.

The PROMOTED flag is cleared every time the node is turned back into a
replica of some other node.
---
 src/cluster.c | 15 ++++++++++++---
 src/redis.h   |  1 +
 2 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/src/cluster.c b/src/cluster.c
index 149b9286f..f26cc98c1 100644
--- a/src/cluster.c
+++ b/src/cluster.c
@@ -115,6 +115,8 @@ int clusterLoadConfig(char *filename) {
                 n->flags |= REDIS_NODE_HANDSHAKE;
             } else if (!strcasecmp(s,"noaddr")) {
                 n->flags |= REDIS_NODE_NOADDR;
+            } else if (!strcasecmp(s,"promoted")) {
+                n->flags |= REDIS_NODE_PROMOTED;
             } else if (!strcasecmp(s,"noflags")) {
                 /* nothing to do */
             } else {
@@ -675,6 +677,7 @@ void clusterProcessGossipSection(clusterMsg *hdr, clusterLink *link) {
         if (flags & REDIS_NODE_FAIL) ci = sdscat(ci,"fail,");
         if (flags & REDIS_NODE_HANDSHAKE) ci = sdscat(ci,"handshake,");
         if (flags & REDIS_NODE_NOADDR) ci = sdscat(ci,"noaddr,");
+        if (flags & REDIS_NODE_PROMOTED) ci = sdscat(ci,"promoted,");
         if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';
 
         redisLog(REDIS_DEBUG,"GOSSIP %.40s %s:%d %s",
@@ -767,6 +770,7 @@ int clusterProcessPacket(clusterLink *link) {
     clusterMsg *hdr = (clusterMsg*) link->rcvbuf;
     uint32_t totlen = ntohl(hdr->totlen);
     uint16_t type = ntohs(hdr->type);
+    uint16_t flags = ntohs(hdr->flags);
     clusterNode *sender;
 
     redisLog(REDIS_DEBUG,"--- Processing packet of type %d, %lu bytes",
@@ -896,10 +900,12 @@ int clusterProcessPacket(clusterLink *link) {
                     sender->flags |= REDIS_NODE_MASTER;
                     sender->slaveof = NULL;
 
-                    /* If this node used to be our slave, it means that
-                     * we were failed over. We'll turn ourself into a slave
+                    /* If this node used to be our slave, and now has the
+                     * PROMOTED flag set. We'll turn ourself into a slave
                      * of the new master. */
-                    if (oldmaster == server.cluster->myself) {
+                    if (flags & REDIS_NODE_PROMOTED &&
+                        oldmaster == server.cluster->myself)
+                    {
                         redisLog(REDIS_WARNING,"One of my slaves took my place. Reconfiguring myself as a replica of %.40s", sender->name);
                         clusterSetMaster(sender);
                     }
@@ -1441,6 +1447,7 @@ void clusterHandleSlaveFailover(void) {
                                server.cluster->myself);
         server.cluster->myself->flags &= ~REDIS_NODE_SLAVE;
         server.cluster->myself->flags |= REDIS_NODE_MASTER;
+        server.cluster->myself->flags |= REDIS_NODE_PROMOTED;
         server.cluster->myself->slaveof = NULL;
         replicationUnsetMaster();
 
@@ -1872,6 +1879,7 @@ sds clusterGenNodesDescription(void) {
         if (node->flags & REDIS_NODE_FAIL) ci = sdscat(ci,"fail,");
         if (node->flags & REDIS_NODE_HANDSHAKE) ci =sdscat(ci,"handshake,");
         if (node->flags & REDIS_NODE_NOADDR) ci = sdscat(ci,"noaddr,");
+        if (node->flags & REDIS_NODE_PROMOTED) ci = sdscat(ci,"promoted,");
         if (ci[sdslen(ci)-1] == ',') ci[sdslen(ci)-1] = ' ';
 
         /* Slave of... or just "-" */
@@ -2236,6 +2244,7 @@ void clusterCommand(redisClient *c) {
         }
 
         /* Set the master. */
+        server.cluster->myself->flags &= ~REDIS_NODE_PROMOTED;
         clusterSetMaster(n);
         clusterUpdateState();
         clusterSaveConfigOrDie();
diff --git a/src/redis.h b/src/redis.h
index 4408925c2..fc6eb95c9 100644
--- a/src/redis.h
+++ b/src/redis.h
@@ -542,6 +542,7 @@ typedef struct clusterLink {
 #define REDIS_NODE_HANDSHAKE 32 /* We have still to exchange the first ping */
 #define REDIS_NODE_NOADDR   64  /* We don't know the address of this node */
 #define REDIS_NODE_MEET 128     /* Send a MEET message to this node */
+#define REDIS_NODE_PROMOTED 256 /* Master was a slave propoted by failover */
 #define REDIS_NODE_NULL_NAME "\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000"
 
 /* This structure represent elements of node->fail_reports. */