Source Version: 2.4.4
Update 2014.3.17, base 2.8.7
The master-slave replication of redis is simple but powerful with the following features:
1. One master supports multiple slave connections, and slaves can accept connections from other slaves
2. Master and slave are non-blocking when master-slave synchronization occurs
redis master-slave replication can be used to:
1. data redundancy
2. slave provides some read-only services as an extension of master
3. You can improve master performance by putting data persistence in slave
By simply configuring slaves (master side does not require configuration), users can use redis master-slave replication
Related Configuration (redis.conf):
slaveof <masterip> <masterport>
Indicates that the redis service is slave, master ip and master ports are master's ip and port, respectively
masterauth <master-password>
If master has set a secure password, set it here as the corresponding password
slave-serve-stale-data yes
When slave loses master or synchronization is in progress, if a service request to slave occurs:
slave-serve-stale-data is set to yes and slave is still serving properly
slave-serve-stale-data set to no then slave returns a client error: "SYNC with master in progress"
repl-ping-slave-period 10
Time interval between slave sending PINGS and master
repl-timeout 60
IO Timeout
Code:
slave end
slave status:
/* Slave replication state - slave side */
#define REDIS_REPL_NONE 0 /* No active replication */
#define REDIS_REPL_CONNECT 1 /* Must connect to master */
#define REDIS_REPL_CONNECTING 2 /* Connecting to master */
#define REDIS_REPL_TRANSFER 3 /* Receiving .rdb from master */
#define REDIS_REPL_CONNECTED 4 /* Connected to master */
Settings at initialization
server.replstate = REDIS_REPL_CONNECT
slave needs to connect master
Slave periodically calls replicationCron to see slave status:
void replicationCron(void) { /*Determine if IO timed out*/ if (server.masterhost && server.replstate == REDIS_REPL_TRANSFER && (time(NULL)-server.repl_transfer_lastio) > server.repl_timeout) { redisLog(REDIS_WARNING,"Timeout receiving bulk data from MASTER..."); replicationAbortSyncTransfer(); //Terminate the connection and set server.replstate = REDIS_REPL_CONNECT; } /* Timed out master when we are an already connected slave? */ if (server.masterhost && server.replstate == REDIS_REPL_CONNECTED && (time(NULL)-server.master->lastinteraction) > server.repl_timeout) { redisLog(REDIS_WARNING,"MASTER time out: no data nor PING received..."); freeClient(server.master); } /* Check if we should connect to a MASTER */ if (server.replstate == REDIS_REPL_CONNECT) { redisLog(REDIS_NOTICE,"Connecting to MASTER..."); if (connectWithMaster() == REDIS_OK) { //Connect master redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started"); } } /* If we have attached slaves, PING them from time to time. * So slaves can implement an explicit timeout to masters, and will * be able to detect a link disconnection even if the TCP connection * will not actually go down. */ if (!(server.cronloops % (server.repl_ping_slave_period*10))) { listIter li; listNode *ln; listRewind(server.slaves,&li); while((ln = listNext(&li))) { redisClient *slave = ln->value; /* Don't ping slaves that are in the middle of a bulk transfer * with the master for first synchronization. */ if (slave->replstate == REDIS_REPL_SEND_BULK) continue; if (slave->replstate == REDIS_REPL_ONLINE) { /* If the slave is online send a normal ping */ addReplySds(slave,sdsnew("PING\r\n")); } else { /* Otherwise we are in the pre-synchronization stage. * Just a newline will do the work of refreshing the * connection last interaction time, and at the same time * we'll be sure that being a single char there are no * short-write problems. */ if (write(slave->fd, "\n", 1) == -1) { /* Don't worry, it's just a ping. */ } } } } }
When server.replstate == REDIS_REPL_CONNECT, slave connects to master. When the connection succeeds, slave executes the syncWithMaster function, and syncWithMaster sends the SYNC command to the master
int connectWithMaster(void) { int fd; fd = anetTcpNonBlockConnect(NULL,server.masterhost,server.masterport); if (fd == -1) { redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s", strerror(errno)); return REDIS_ERR; } if (aeCreateFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE,syncWithMaster,NULL) == AE_ERR) { close(fd); redisLog(REDIS_WARNING,"Can't create readable event for SYNC"); return REDIS_ERR; } server.repl_transfer_s = fd; server.replstate = REDIS_REPL_CONNECTING; return REDIS_OK; }
master side:
master unifies the processing of slave connections and client connections. After receiving the SYNC command issued by slave, syncCommand executes syncCommand, which will see the current status. If a snapshot is being taken, wait, or start a background process to take a snapshot.
void syncCommand(redisClient *c) { /* ignore SYNC if aleady slave or in monitor mode */ if (c->flags & REDIS_SLAVE) return; /* Refuse SYNC requests if we are a slave but the link with our master * is not ok... */ if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED) { addReplyError(c,"Can't SYNC while not connected with my master"); return; } /* SYNC can't be issued when the server has pending data to send to * the client about already issued commands. We need a fresh reply * buffer registering the differences between the BGSAVE and the current * dataset, so that we can copy to other slaves if needed. */ if (listLength(c->reply) != 0) { addReplyError(c,"SYNC is invalid with pending input"); return; } redisLog(REDIS_NOTICE,"Slave ask for synchronization"); /* Here we need to check if there is a background saving operation * in progress, or if it is required to start one */ if (server.bgsavechildpid != -1) { ..... } else { /* Ok we don't have a BGSAVE in progress, let's start one */ redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC"); if (rdbSaveBackground(server.dbfilename) != 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->repldbfd = -1; c->flags |= REDIS_SLAVE; c->slaveseldb = 0; listAddNodeTail(server.slaves,c); return; }
After the snapshot is completed, the updateSlavesWaitingBgsave function is executed, and updateSlavesWaitingBgsave looks at the status of each slave of the current master. If it finds that there is one waiting for bgsave to complete, the registration event sendBulkToSlave, sendBulkToSlave sends the snapshot file to slave
To avoid blocking applications, transfer only 16K data at a timevoid updateSlavesWaitingBgsave(int bgsaveerr) { listNode *ln; int startbgsave = 0; listIter li; listRewind(server.slaves,&li); while((ln = listNext(&li))) { redisClient *slave = ln->value; if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) { startbgsave = 1; slave->replstate = REDIS_REPL_WAIT_BGSAVE_END; } else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) { struct redis_stat buf; if (bgsaveerr != REDIS_OK) { freeClient(slave); redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error"); continue; } if ((slave->repldbfd = open(server.dbfilename,O_RDONLY)) == -1 || redis_fstat(slave->repldbfd,&buf) == -1) { freeClient(slave); redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno)); continue; } slave->repldboff = 0; slave->repldbsize = buf.st_size; slave->replstate = REDIS_REPL_SEND_BULK; aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE); //Delete previous callbacks if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) { //Register new write callbacks freeClient(slave); continue; } } } if (startbgsave) { if (rdbSaveBackground(server.dbfilename) != REDIS_OK) { listIter li; listRewind(server.slaves,&li); redisLog(REDIS_WARNING,"SYNC failed. BGSAVE failed"); while((ln = listNext(&li))) { redisClient *slave = ln->value; if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) freeClient(slave); } } } }
void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) { ...... lseek(slave->repldbfd,slave->repldboff,SEEK_SET); //Pointer moves to last sent position buflen = read(slave->repldbfd,buf,REDIS_IOBUF_LEN); //Read 16K data ...... if ((nwritten = write(fd,buf,buflen)) == -1) { //Transfer data to slave if (errno != EAGAIN) { redisLog(REDIS_WARNING,"Write error sending DB to slave: %s", strerror(errno)); freeClient(slave); } return; } slave->repldboff += nwritten; //Update Sent Location ...... }
After slave completes its first synchronization, subsequently if master receives commands to change db state, call replicationFeedSlaves to send the corresponding changes to slave
/* Call() is the core of Redis execution of a command */ void call(redisClient *c) { long long dirty, start = ustime(), duration; dirty = server.dirty; c->cmd->proc(c); dirty = server.dirty-dirty; duration = ustime()-start; slowlogPushEntryIfNeeded(c->argv,c->argc,duration); if (server.appendonly && dirty > 0) feedAppendOnlyFile(c->cmd,c->db->id,c->argv,c->argc); if ((dirty > 0 || c->cmd->flags & REDIS_CMD_FORCE_REPLICATION) && listLength(server.slaves)) replicationFeedSlaves(server.slaves,c->db->id,c->argv,c->argc); if (listLength(server.monitors)) replicationFeedMonitors(server.monitors,c->db->id,c->argv,c->argc); server.stat_numcommands++; }
Summary:
1. redis master-slave replication does not add much extra code, but it is powerful, supports multiple slaves, and supports slaves as masters.
2. redis claims that master-slave replication is non-blocking, but since redis uses a single-threaded service and the interaction with slave is handled by the processing threads in unison, it has a performance impact
Reprinted at: https://my.oschina.net/zipu888/blog/549581