From a3309139992de03f3e68ec648db188c1c43057a0 Mon Sep 17 00:00:00 2001 From: Pieter Noordhuis Date: Thu, 19 May 2011 18:53:06 +0200 Subject: [PATCH 1/3] Non-blocking connect with master --- src/redis.h | 10 +++--- src/replication.c | 82 ++++++++++++++++++++++++++++++----------------- 2 files changed, 58 insertions(+), 34 deletions(-) diff --git a/src/redis.h b/src/redis.h index 5934b6a6..d6b99cea 100644 --- a/src/redis.h +++ b/src/redis.h @@ -153,10 +153,11 @@ #define REDIS_REQ_MULTIBULK 2 /* 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_TRANSFER 2 /* Receiving .rdb from master */ -#define REDIS_REPL_CONNECTED 3 /* Connected to master */ +#define REDIS_REPL_NONE 0 /* No active replication */ +#define REDIS_REPL_CONNECT 1 /* Must connect to master */ +#define REDIS_REPL_TRANSFER 2 /* Receiving .rdb from master */ +#define REDIS_REPL_CONNECTED 3 /* Connected to master */ +#define REDIS_REPL_CONNECTING 4 /* Connecting to master */ /* Slave replication state - from the point of view of master * Note that in SEND_BULK and ONLINE state the slave receives new updates @@ -888,7 +889,6 @@ int fwriteBulkCount(FILE *fp, char prefix, int count); /* Replication */ void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc); void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc); -int syncWithMaster(void); void updateSlavesWaitingBgsave(int bgsaveerr); void replicationCron(void); diff --git a/src/replication.c b/src/replication.c index b0fa7055..c3ddd2d3 100644 --- a/src/replication.c +++ b/src/replication.c @@ -393,46 +393,44 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { } } -int syncWithMaster(void) { - char buf[1024], tmpfile[256], authcmd[1024]; - int fd = anetTcpConnect(NULL,server.masterhost,server.masterport); +void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) { + char buf[1024], tmpfile[256]; int dfd, maxtries = 5; + REDIS_NOTUSED(el); + REDIS_NOTUSED(privdata); + REDIS_NOTUSED(mask); - if (fd == -1) { - redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s", - strerror(errno)); - return REDIS_ERR; - } + /* Remove this event before anything else. */ + aeDeleteFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE); /* AUTH with the master if required. */ if(server.masterauth) { - snprintf(authcmd, 1024, "AUTH %s\r\n", server.masterauth); - if (syncWrite(fd, authcmd, strlen(server.masterauth)+7, 5) == -1) { - close(fd); + char authcmd[1024]; + size_t authlen; + + authlen = snprintf(authcmd,sizeof(authcmd),"AUTH %s\r\n",server.masterauth); + if (syncWrite(fd,authcmd,authlen,5) == -1) { redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s", strerror(errno)); - return REDIS_ERR; - } + goto error; + } /* Read the AUTH result. */ if (syncReadLine(fd,buf,1024,3600) == -1) { - close(fd); redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s", strerror(errno)); - return REDIS_ERR; + goto error; } if (buf[0] != '+') { - close(fd); redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?"); - return REDIS_ERR; + goto error; } } /* Issue the SYNC command */ if (syncWrite(fd,"SYNC \r\n",7,5) == -1) { - close(fd); redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s", strerror(errno)); - return REDIS_ERR; + goto error; } /* Prepare a suitable temp file for bulk transfer */ @@ -444,25 +442,51 @@ int syncWithMaster(void) { sleep(1); } if (dfd == -1) { - close(fd); redisLog(REDIS_WARNING,"Opening the temp file needed for MASTER <-> SLAVE synchronization: %s",strerror(errno)); - return REDIS_ERR; + goto error; } /* Setup the non blocking download of the bulk file. */ - if (aeCreateFileEvent(server.el, fd, AE_READABLE, readSyncBulkPayload, NULL) + if (aeCreateFileEvent(server.el,fd, AE_READABLE,readSyncBulkPayload,NULL) == AE_ERR) + { + redisLog(REDIS_WARNING,"Can't create readable event for SYNC"); + goto error; + } + + server.replstate = REDIS_REPL_TRANSFER; + server.repl_transfer_left = -1; + server.repl_transfer_fd = dfd; + server.repl_transfer_lastio = time(NULL); + server.repl_transfer_tmpfile = zstrdup(tmpfile); + return; + +error: + server.replstate = REDIS_REPL_CONNECT; + close(fd); + return; +} + +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.replstate = REDIS_REPL_TRANSFER; - server.repl_transfer_left = -1; + server.repl_transfer_s = fd; - server.repl_transfer_fd = dfd; - server.repl_transfer_lastio = time(NULL); - server.repl_transfer_tmpfile = zstrdup(tmpfile); + server.replstate = REDIS_REPL_CONNECTING; return REDIS_OK; } @@ -517,8 +541,8 @@ void replicationCron(void) { /* Check if we should connect to a MASTER */ if (server.replstate == REDIS_REPL_CONNECT) { redisLog(REDIS_NOTICE,"Connecting to MASTER..."); - if (syncWithMaster() == REDIS_OK) { - redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started: SYNC sent"); + if (connectWithMaster() == REDIS_OK) { + redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started"); if (server.appendonly) rewriteAppendOnlyFileBackground(); } } From b075621fb7d8b44b10d849e8db6db9b2498056ad Mon Sep 17 00:00:00 2001 From: Pieter Noordhuis Date: Sun, 22 May 2011 12:41:24 +0200 Subject: [PATCH 2/3] Minor changes in non-blocking repl. connect --- src/redis.h | 6 +++--- src/replication.c | 29 +++++++++++++++++------------ 2 files changed, 20 insertions(+), 15 deletions(-) diff --git a/src/redis.h b/src/redis.h index d6b99cea..19f0b516 100644 --- a/src/redis.h +++ b/src/redis.h @@ -155,9 +155,9 @@ /* 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_TRANSFER 2 /* Receiving .rdb from master */ -#define REDIS_REPL_CONNECTED 3 /* Connected to master */ -#define REDIS_REPL_CONNECTING 4 /* Connecting 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 */ /* Slave replication state - from the point of view of master * Note that in SEND_BULK and ONLINE state the slave receives new updates diff --git a/src/replication.c b/src/replication.c index c3ddd2d3..851a40f8 100644 --- a/src/replication.c +++ b/src/replication.c @@ -319,15 +319,14 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { redisLog(REDIS_WARNING, "I/O error reading bulk count from MASTER: %s", strerror(errno)); - replicationAbortSyncTransfer(); - return; + goto error; } + if (buf[0] == '-') { redisLog(REDIS_WARNING, "MASTER aborted replication with an error: %s", buf+1); - replicationAbortSyncTransfer(); - return; + goto error; } else if (buf[0] == '\0') { /* At this stage just a newline works as a PING in order to take * the connection live. So we refresh our last interaction @@ -336,8 +335,7 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { return; } else if (buf[0] != '$') { redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$', are you sure the host and port are right?"); - replicationAbortSyncTransfer(); - return; + goto error; } server.repl_transfer_left = strtol(buf+1,NULL,10); redisLog(REDIS_NOTICE, @@ -359,8 +357,7 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { server.repl_transfer_lastio = time(NULL); if (write(server.repl_transfer_fd,buf,nread) != nread) { redisLog(REDIS_WARNING,"Write error or short write writing to the DB dump file needed for MASTER <-> SLAVE synchrnonization: %s", strerror(errno)); - replicationAbortSyncTransfer(); - return; + goto error; } server.repl_transfer_left -= nread; /* Check if the transfer is now complete */ @@ -391,6 +388,12 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { server.replstate = REDIS_REPL_CONNECTED; redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Finished with success"); } + + return; + +error: + replicationAbortSyncTransfer(); + return; } void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) { @@ -400,8 +403,10 @@ void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) { REDIS_NOTUSED(privdata); REDIS_NOTUSED(mask); - /* Remove this event before anything else. */ - aeDeleteFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE); + /* This event should only be triggered once since it is used to have a + * non-blocking connect(2) to the master. It has been triggered when this + * function is called, so we can delete it. */ + aeDeleteFileEvent(server.el,fd,AE_WRITABLE); /* AUTH with the master if required. */ if(server.masterauth) { @@ -477,8 +482,8 @@ int connectWithMaster(void) { return REDIS_ERR; } - if (aeCreateFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE, - syncWithMaster,NULL) == AE_ERR) + if (aeCreateFileEvent(server.el,fd,AE_WRITABLE,syncWithMaster,NULL) == + AE_ERR) { close(fd); redisLog(REDIS_WARNING,"Can't create readable event for SYNC"); From 890a2ed989274cb09b5cde1def3935e110ec3cb9 Mon Sep 17 00:00:00 2001 From: Pieter Noordhuis Date: Sun, 22 May 2011 12:57:30 +0200 Subject: [PATCH 3/3] Configurable synchronous I/O timeout --- src/redis.c | 1 + src/redis.h | 4 ++++ src/replication.c | 8 ++++---- 3 files changed, 9 insertions(+), 4 deletions(-) diff --git a/src/redis.c b/src/redis.c index 63b41ba8..f64a2d34 100644 --- a/src/redis.c +++ b/src/redis.c @@ -871,6 +871,7 @@ void initServerConfig() { server.masterport = 6379; server.master = NULL; server.replstate = REDIS_REPL_NONE; + server.repl_syncio_timeout = REDIS_REPL_SYNCIO_TIMEOUT; server.repl_serve_stale_data = 1; /* Double constants initialization */ diff --git a/src/redis.h b/src/redis.h index 19f0b516..f249d237 100644 --- a/src/redis.h +++ b/src/redis.h @@ -159,6 +159,9 @@ #define REDIS_REPL_TRANSFER 3 /* Receiving .rdb from master */ #define REDIS_REPL_CONNECTED 4 /* Connected to master */ +/* Synchronous read timeout - slave side */ +#define REDIS_REPL_SYNCIO_TIMEOUT 5 + /* Slave replication state - from the point of view of master * Note that in SEND_BULK and ONLINE state the slave receives new updates * in its output queue. In the WAIT_BGSAVE state instead the server is waiting @@ -586,6 +589,7 @@ struct redisServer { char *masterhost; int masterport; redisClient *master; /* client that is master for this slave */ + int repl_syncio_timeout; /* timeout for synchronous I/O calls */ int replstate; /* replication status if the instance is a slave */ off_t repl_transfer_left; /* bytes left reading .rdb */ int repl_transfer_s; /* slave -> master SYNC socket */ diff --git a/src/replication.c b/src/replication.c index 851a40f8..87d575ca 100644 --- a/src/replication.c +++ b/src/replication.c @@ -315,7 +315,7 @@ void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) { /* If repl_transfer_left == -1 we still have to read the bulk length * from the master reply. */ if (server.repl_transfer_left == -1) { - if (syncReadLine(fd,buf,1024,3600) == -1) { + if (syncReadLine(fd,buf,1024,server.repl_syncio_timeout) == -1) { redisLog(REDIS_WARNING, "I/O error reading bulk count from MASTER: %s", strerror(errno)); @@ -414,13 +414,13 @@ void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) { size_t authlen; authlen = snprintf(authcmd,sizeof(authcmd),"AUTH %s\r\n",server.masterauth); - if (syncWrite(fd,authcmd,authlen,5) == -1) { + if (syncWrite(fd,authcmd,authlen,server.repl_syncio_timeout) == -1) { redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s", strerror(errno)); goto error; } /* Read the AUTH result. */ - if (syncReadLine(fd,buf,1024,3600) == -1) { + if (syncReadLine(fd,buf,1024,server.repl_syncio_timeout) == -1) { redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s", strerror(errno)); goto error; @@ -432,7 +432,7 @@ void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) { } /* Issue the SYNC command */ - if (syncWrite(fd,"SYNC \r\n",7,5) == -1) { + if (syncWrite(fd,"SYNC \r\n",7,server.repl_syncio_timeout) == -1) { redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s", strerror(errno)); goto error;