diff options
author | Kevin McGehee <mcgehee@amazon.com> | 2015-10-14 12:03:47 -0700 |
---|---|---|
committer | antirez <antirez@gmail.com> | 2015-10-15 10:22:13 +0200 |
commit | dc03e4c51b6ec1cccaaa0a76f61832e499f5871b (patch) | |
tree | d2ed2061652840fb1cc9fe32360d062874e2bdd5 | |
parent | 30978004b33575bb9c1ff45ff8df166792a90f8b (diff) | |
download | redis-dc03e4c51b6ec1cccaaa0a76f61832e499f5871b.tar.gz |
Fix master timeout during handshake
This change allows a slave to properly time out a dead master during
the extended asynchronous synchronization state machine. Now, slaves
will record their last interaction with the master and apply the
replication timeout before a response to the PSYNC request is received.
-rw-r--r-- | src/replication.c | 22 |
1 files changed, 19 insertions, 3 deletions
diff --git a/src/replication.c b/src/replication.c index cef65dfd7..e6b64d729 100644 --- a/src/replication.c +++ b/src/replication.c @@ -41,6 +41,7 @@ void replicationDiscardCachedMaster(void); void replicationResurrectCachedMaster(int newfd); void replicationSendAck(void); void putSlaveOnline(redisClient *slave); +int serverInHandshakeState(int repl_state); /* --------------------------- Utility functions ---------------------------- */ @@ -1196,6 +1197,7 @@ char *sendSynchronousCommand(int flags, int fd, ...) { return sdscatprintf(sdsempty(),"-Reading from master: %s", strerror(errno)); } + server.repl_transfer_lastio = server.unixtime; return sdsnew(buf); } return NULL; @@ -1626,7 +1628,7 @@ void undoConnectWithMaster(void) { int fd = server.repl_transfer_s; redisAssert(server.repl_state == REDIS_REPL_CONNECTING || - server.repl_state == REDIS_REPL_RECEIVE_PONG); + serverInHandshakeState(server.repl_state)); aeDeleteFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE); close(fd); server.repl_transfer_s = -1; @@ -1645,7 +1647,7 @@ int cancelReplicationHandshake(void) { if (server.repl_state == REDIS_REPL_TRANSFER) { replicationAbortSyncTransfer(); } else if (server.repl_state == REDIS_REPL_CONNECTING || - server.repl_state == REDIS_REPL_RECEIVE_PONG) + serverInHandshakeState(server.repl_state)) { undoConnectWithMaster(); } else { @@ -1802,6 +1804,20 @@ void roleCommand(redisClient *c) { } } +/* Returns 1 if the given replication state is a handshake state, + * 0 otherwise. */ +int serverInHandshakeState(int repl_state) { + return repl_state == REDIS_REPL_RECEIVE_PONG || + repl_state == REDIS_REPL_SEND_AUTH || + repl_state == REDIS_REPL_RECEIVE_AUTH || + repl_state == REDIS_REPL_SEND_PORT || + repl_state == REDIS_REPL_RECEIVE_PORT || + repl_state == REDIS_REPL_SEND_CAPA || + repl_state == REDIS_REPL_RECEIVE_CAPA || + repl_state == REDIS_REPL_SEND_PSYNC || + repl_state == REDIS_REPL_RECEIVE_PSYNC; +} + /* Send a REPLCONF ACK command to the master to inform it about the current * processed offset. If we are not connected with a master, the command has * no effects. */ @@ -2186,7 +2202,7 @@ void replicationCron(void) { /* Non blocking connection timeout? */ if (server.masterhost && (server.repl_state == REDIS_REPL_CONNECTING || - server.repl_state == REDIS_REPL_RECEIVE_PONG) && + serverInHandshakeState(server.repl_state)) && (time(NULL)-server.repl_transfer_lastio) > server.repl_timeout) { redisLog(REDIS_WARNING,"Timeout connecting to the MASTER..."); |