[redis.git] / src / replication.c

#include "redis.h"

#include <sys/time.h>
#include <unistd.h>
#include <fcntl.h>
#include <sys/stat.h>

/* ---------------------------------- MASTER -------------------------------- */

void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) {
    listNode *ln;
    listIter li;
    int outc = 0, j;
    robj **outv;
    /* We need 1+(ARGS*3) objects since commands are using the new protocol
     * and we one 1 object for the first "*<count>\r\n" multibulk count, then
     * for every additional object we have "$<count>\r\n" + object + "\r\n". */
    robj *static_outv[REDIS_STATIC_ARGS*3+1];
    robj *lenobj;

    if (argc <= REDIS_STATIC_ARGS) {
        outv = static_outv;
    } else {
        outv = zmalloc(sizeof(robj*)*(argc*3+1));
    }

    lenobj = createObject(REDIS_STRING,
            sdscatprintf(sdsempty(), "*%d\r\n", argc));
    lenobj->refcount = 0;
    outv[outc++] = lenobj;
    for (j = 0; j < argc; j++) {
        lenobj = createObject(REDIS_STRING,
            sdscatprintf(sdsempty(),"$%lu\r\n",
                (unsigned long) stringObjectLen(argv[j])));
        lenobj->refcount = 0;
        outv[outc++] = lenobj;
        outv[outc++] = argv[j];
        outv[outc++] = shared.crlf;
    }

    /* Increment all the refcounts at start and decrement at end in order to
     * be sure to free objects if there is no slave in a replication state
     * able to be feed with commands */
    for (j = 0; j < outc; j++) incrRefCount(outv[j]);
    listRewind(slaves,&li);
    while((ln = listNext(&li))) {
        redisClient *slave = ln->value;

        /* Don't feed slaves that are still waiting for BGSAVE to start */
        if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) continue;

        /* Feed all the other slaves, MONITORs and so on */
        if (slave->slaveseldb != dictid) {
            robj *selectcmd;

            switch(dictid) {
            case 0: selectcmd = shared.select0; break;
            case 1: selectcmd = shared.select1; break;
            case 2: selectcmd = shared.select2; break;
            case 3: selectcmd = shared.select3; break;
            case 4: selectcmd = shared.select4; break;
            case 5: selectcmd = shared.select5; break;
            case 6: selectcmd = shared.select6; break;
            case 7: selectcmd = shared.select7; break;
            case 8: selectcmd = shared.select8; break;
            case 9: selectcmd = shared.select9; break;
            default:
                selectcmd = createObject(REDIS_STRING,
                    sdscatprintf(sdsempty(),"select %d\r\n",dictid));
                selectcmd->refcount = 0;
                break;
            }
            addReply(slave,selectcmd);
            slave->slaveseldb = dictid;
        }
        for (j = 0; j < outc; j++) addReply(slave,outv[j]);
    }
    for (j = 0; j < outc; j++) decrRefCount(outv[j]);
    if (outv != static_outv) zfree(outv);
}

void replicationFeedMonitors(list *monitors, int dictid, robj **argv, int argc) {
    listNode *ln;
    listIter li;
    int j;
    sds cmdrepr = sdsnew("+");
    robj *cmdobj;
    struct timeval tv;

    gettimeofday(&tv,NULL);
    cmdrepr = sdscatprintf(cmdrepr,"%ld.%06ld ",(long)tv.tv_sec,(long)tv.tv_usec);
    if (dictid != 0) cmdrepr = sdscatprintf(cmdrepr,"(db %d) ", dictid);

    for (j = 0; j < argc; j++) {
        if (argv[j]->encoding == REDIS_ENCODING_INT) {
            cmdrepr = sdscatprintf(cmdrepr, "\"%ld\"", (long)argv[j]->ptr);
        } else {
            cmdrepr = sdscatrepr(cmdrepr,(char*)argv[j]->ptr,
                        sdslen(argv[j]->ptr));
        }
        if (j != argc-1)
            cmdrepr = sdscatlen(cmdrepr," ",1);
    }
    cmdrepr = sdscatlen(cmdrepr,"\r\n",2);
    cmdobj = createObject(REDIS_STRING,cmdrepr);

    listRewind(monitors,&li);
    while((ln = listNext(&li))) {
        redisClient *monitor = ln->value;
        addReply(monitor,cmdobj);
    }
    decrRefCount(cmdobj);
}

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) {
        /* Ok a background save is in progress. Let's check if it is a good
         * one for replication, i.e. if there is another slave that is
         * registering differences since the server forked to save */
        redisClient *slave;
        listNode *ln;
        listIter li;

        listRewind(server.slaves,&li);
        while((ln = listNext(&li))) {
            slave = ln->value;
            if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) break;
        }
        if (ln) {
            /* Perfect, the server is already registering differences for
             * another slave. Set the right state, and copy the buffer. */
            listRelease(c->reply);
            c->reply = listDup(slave->reply);
            c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
            redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC");
        } else {
            /* No way, we need to wait for the next BGSAVE in order to
             * register differences */
            c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
            redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
        }
    } 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;
}

void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) {
    redisClient *slave = privdata;
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(mask);
    char buf[REDIS_IOBUF_LEN];
    ssize_t nwritten, buflen;

    if (slave->repldboff == 0) {
        /* Write the bulk write count before to transfer the DB. In theory here
         * we don't know how much room there is in the output buffer of the
         * socket, but in pratice SO_SNDLOWAT (the minimum count for output
         * operations) will never be smaller than the few bytes we need. */
        sds bulkcount;

        bulkcount = sdscatprintf(sdsempty(),"$%lld\r\n",(unsigned long long)
            slave->repldbsize);
        if (write(fd,bulkcount,sdslen(bulkcount)) != (signed)sdslen(bulkcount))
        {
            sdsfree(bulkcount);
            freeClient(slave);
            return;
        }
        sdsfree(bulkcount);
    }
    lseek(slave->repldbfd,slave->repldboff,SEEK_SET);
    buflen = read(slave->repldbfd,buf,REDIS_IOBUF_LEN);
    if (buflen <= 0) {
        redisLog(REDIS_WARNING,"Read error sending DB to slave: %s",
            (buflen == 0) ? "premature EOF" : strerror(errno));
        freeClient(slave);
        return;
    }
    if ((nwritten = write(fd,buf,buflen)) == -1) {
        redisLog(REDIS_VERBOSE,"Write error sending DB to slave: %s",
            strerror(errno));
        freeClient(slave);
        return;
    }
    slave->repldboff += nwritten;
    if (slave->repldboff == slave->repldbsize) {
        close(slave->repldbfd);
        slave->repldbfd = -1;
        aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
        slave->replstate = REDIS_REPL_ONLINE;
        if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
            sendReplyToClient, slave) == AE_ERR) {
            freeClient(slave);
            return;
        }
        addReplySds(slave,sdsempty());
        redisLog(REDIS_NOTICE,"Synchronization with slave succeeded");
    }
}

/* This function is called at the end of every backgrond saving.
 * The argument bgsaveerr is REDIS_OK if the background saving succeeded
 * otherwise REDIS_ERR is passed to the function.
 *
 * The goal of this function is to handle slaves waiting for a successful
 * background saving in order to perform non-blocking synchronization. */
void 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);
            if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
                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);
            }
        }
    }
}

/* ----------------------------------- SLAVE -------------------------------- */

/* Abort the async download of the bulk dataset while SYNC-ing with master */
void replicationAbortSyncTransfer(void) {
    redisAssert(server.replstate == REDIS_REPL_TRANSFER);

    aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
    close(server.repl_transfer_s);
    close(server.repl_transfer_fd);
    unlink(server.repl_transfer_tmpfile);
    zfree(server.repl_transfer_tmpfile);
    server.replstate = REDIS_REPL_CONNECT;
}

/* Asynchronously read the SYNC payload we receive from a master */
void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
    char buf[4096];
    ssize_t nread, readlen;
    REDIS_NOTUSED(el);
    REDIS_NOTUSED(privdata);
    REDIS_NOTUSED(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) {
            redisLog(REDIS_WARNING,
                "I/O error reading bulk count from MASTER: %s",
                strerror(errno));
            replicationAbortSyncTransfer();
            return;
        }
        if (buf[0] == '-') {
            redisLog(REDIS_WARNING,
                "MASTER aborted replication with an error: %s",
                buf+1);
            replicationAbortSyncTransfer();
            return;
        } 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
             * timestamp. */
            server.repl_transfer_lastio = time(NULL);
            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;
        }
        server.repl_transfer_left = strtol(buf+1,NULL,10);
        redisLog(REDIS_NOTICE,
            "MASTER <-> SLAVE sync: receiving %ld bytes from master",
            server.repl_transfer_left);
        return;
    }

    /* Read bulk data */
    readlen = (server.repl_transfer_left < (signed)sizeof(buf)) ?
        server.repl_transfer_left : (signed)sizeof(buf);
    nread = read(fd,buf,readlen);
    if (nread <= 0) {
        redisLog(REDIS_WARNING,"I/O error trying to sync with MASTER: %s",
            (nread == -1) ? strerror(errno) : "connection lost");
        replicationAbortSyncTransfer();
        return;
    }
    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;
    }
    server.repl_transfer_left -= nread;
    /* Check if the transfer is now complete */
    if (server.repl_transfer_left == 0) {
        if (rename(server.repl_transfer_tmpfile,server.dbfilename) == -1) {
            redisLog(REDIS_WARNING,"Failed trying to rename the temp DB into dump.rdb in MASTER <-> SLAVE synchronization: %s", strerror(errno));
            replicationAbortSyncTransfer();
            return;
        }
        redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Loading DB in memory");
        emptyDb();
        /* Before loading the DB into memory we need to delete the readable
         * handler, otherwise it will get called recursively since
         * rdbLoad() will call the event loop to process events from time to
         * time for non blocking loading. */
        aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
        if (rdbLoad(server.dbfilename) != REDIS_OK) {
            redisLog(REDIS_WARNING,"Failed trying to load the MASTER synchronization DB from disk");
            replicationAbortSyncTransfer();
            return;
        }
        /* Final setup of the connected slave <- master link */
        zfree(server.repl_transfer_tmpfile);
        close(server.repl_transfer_fd);
        server.master = createClient(server.repl_transfer_s);
        server.master->flags |= REDIS_MASTER;
        server.master->authenticated = 1;
        server.replstate = REDIS_REPL_CONNECTED;
        redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Finished with success");
    }
}

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);

    /* Remove this event before anything else. */
    aeDeleteFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE);

    /* AUTH with the master if required. */
    if(server.masterauth) {
        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));
            goto error;
        }
        /* Read the AUTH result.  */
        if (syncReadLine(fd,buf,1024,3600) == -1) {
            redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s",
                strerror(errno));
            goto error;
        }
        if (buf[0] != '+') {
            redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?");
            goto error;
        }
    }

    /* Issue the SYNC command */
    if (syncWrite(fd,"SYNC \r\n",7,5) == -1) {
        redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s",
            strerror(errno));
        goto error;
    }

    /* Prepare a suitable temp file for bulk transfer */
    while(maxtries--) {
        snprintf(tmpfile,256,
            "temp-%d.%ld.rdb",(int)time(NULL),(long int)getpid());
        dfd = open(tmpfile,O_CREAT|O_WRONLY|O_EXCL,0644);
        if (dfd != -1) break;
        sleep(1);
    }
    if (dfd == -1) {
        redisLog(REDIS_WARNING,"Opening the temp file needed for MASTER <-> SLAVE synchronization: %s",strerror(errno));
        goto error;
    }

    /* Setup the non blocking download of the bulk file. */
    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.repl_transfer_s = fd;
    server.replstate = REDIS_REPL_CONNECTING;
    return REDIS_OK;
}

void slaveofCommand(redisClient *c) {
    if (!strcasecmp(c->argv[1]->ptr,"no") &&
        !strcasecmp(c->argv[2]->ptr,"one")) {
        if (server.masterhost) {
            sdsfree(server.masterhost);
            server.masterhost = NULL;
            if (server.master) freeClient(server.master);
            if (server.replstate == REDIS_REPL_TRANSFER)
                replicationAbortSyncTransfer();
            server.replstate = REDIS_REPL_NONE;
            redisLog(REDIS_NOTICE,"MASTER MODE enabled (user request)");
        }
    } else {
        sdsfree(server.masterhost);
        server.masterhost = sdsdup(c->argv[1]->ptr);
        server.masterport = atoi(c->argv[2]->ptr);
        if (server.master) freeClient(server.master);
        if (server.replstate == REDIS_REPL_TRANSFER)
            replicationAbortSyncTransfer();
        server.replstate = REDIS_REPL_CONNECT;
        redisLog(REDIS_NOTICE,"SLAVE OF %s:%d enabled (user request)",
            server.masterhost, server.masterport);
    }
    addReply(c,shared.ok);
}

/* --------------------------- REPLICATION CRON  ---------------------------- */

#define REDIS_REPL_TIMEOUT 60
#define REDIS_REPL_PING_SLAVE_PERIOD 10

void replicationCron(void) {
    /* Bulk transfer I/O timeout? */
    if (server.masterhost && server.replstate == REDIS_REPL_TRANSFER &&
        (time(NULL)-server.repl_transfer_lastio) > REDIS_REPL_TIMEOUT)
    {
        redisLog(REDIS_WARNING,"Timeout receiving bulk data from MASTER...");
        replicationAbortSyncTransfer();
    }

    /* Timed out master when we are an already connected slave? */
    if (server.masterhost && server.replstate == REDIS_REPL_CONNECTED &&
        (time(NULL)-server.master->lastinteraction) > REDIS_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) {
            redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started");
            if (server.appendonly) rewriteAppendOnlyFileBackground();
        }
    }
    
    /* 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 % (REDIS_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. */
                }
            }
        }
    }
}
Commit	Line	Data
e2641e09	1	#include "redis.h"
	2
	3	#include <sys/time.h>
	4	#include <unistd.h>
	5	#include <fcntl.h>
	6	#include <sys/stat.h>
	7
f4aa600b	8	/* ---------------------------------- MASTER -------------------------------- */
f4aa600b	9
e2641e09	10	void replicationFeedSlaves(list slaves, int dictid, robj *argv, int argc) {
	11	listNode *ln;
	12	listIter li;
	13	int outc = 0, j;
	14	robj **outv;
	15	/* We need 1+(ARGS*3) objects since commands are using the new protocol
	16	* and we one 1 object for the first "*<count>\r\n" multibulk count, then
	17	* for every additional object we have "$<count>\r\n" + object + "\r\n". */
	18	robj static_outv[REDIS_STATIC_ARGS3+1];
	19	robj *lenobj;
	20
	21	if (argc <= REDIS_STATIC_ARGS) {
	22	outv = static_outv;
	23	} else {
	24	outv = zmalloc(sizeof(robj)(argc*3+1));
	25	}
	26
	27	lenobj = createObject(REDIS_STRING,
	28	sdscatprintf(sdsempty(), "*%d\r\n", argc));
	29	lenobj->refcount = 0;
	30	outv[outc++] = lenobj;
	31	for (j = 0; j < argc; j++) {
	32	lenobj = createObject(REDIS_STRING,
	33	sdscatprintf(sdsempty(),"$%lu\r\n",
	34	(unsigned long) stringObjectLen(argv[j])));
	35	lenobj->refcount = 0;
	36	outv[outc++] = lenobj;
	37	outv[outc++] = argv[j];
	38	outv[outc++] = shared.crlf;
	39	}
	40
	41	/* Increment all the refcounts at start and decrement at end in order to
	42	* be sure to free objects if there is no slave in a replication state
	43	* able to be feed with commands */
	44	for (j = 0; j < outc; j++) incrRefCount(outv[j]);
	45	listRewind(slaves,&li);
	46	while((ln = listNext(&li))) {
	47	redisClient *slave = ln->value;
	48
	49	/* Don't feed slaves that are still waiting for BGSAVE to start */
	50	if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) continue;
	51
	52	/* Feed all the other slaves, MONITORs and so on */
	53	if (slave->slaveseldb != dictid) {
	54	robj *selectcmd;
	55
	56	switch(dictid) {
	57	case 0: selectcmd = shared.select0; break;
	58	case 1: selectcmd = shared.select1; break;
	59	case 2: selectcmd = shared.select2; break;
	60	case 3: selectcmd = shared.select3; break;
	61	case 4: selectcmd = shared.select4; break;
	62	case 5: selectcmd = shared.select5; break;
	63	case 6: selectcmd = shared.select6; break;
	64	case 7: selectcmd = shared.select7; break;
	65	case 8: selectcmd = shared.select8; break;
	66	case 9: selectcmd = shared.select9; break;
	67	default:
	68	selectcmd = createObject(REDIS_STRING,
	69	sdscatprintf(sdsempty(),"select %d\r\n",dictid));
	70	selectcmd->refcount = 0;
	71	break;
	72	}
	73	addReply(slave,selectcmd);
74	slave->slaveseldb = dictid;
75	}
76	for (j = 0; j < outc; j++) addReply(slave,outv[j]);
77	}
78	for (j = 0; j < outc; j++) decrRefCount(outv[j]);
79	if (outv != static_outv) zfree(outv);
80	}
81
82	void replicationFeedMonitors(list monitors, int dictid, robj *argv, int argc) {
83	listNode *ln;
84	listIter li;
85	int j;
86	sds cmdrepr = sdsnew("+");
87	robj *cmdobj;
88	struct timeval tv;
89
90	gettimeofday(&tv,NULL);
2b2eca1f	91	cmdrepr = sdscatprintf(cmdrepr,"%ld.%06ld ",(long)tv.tv_sec,(long)tv.tv_usec);
e2641e09	92	if (dictid != 0) cmdrepr = sdscatprintf(cmdrepr,"(db %d) ", dictid);
	93
	94	for (j = 0; j < argc; j++) {
	95	if (argv[j]->encoding == REDIS_ENCODING_INT) {
d3b958c3	96	cmdrepr = sdscatprintf(cmdrepr, "\"%ld\"", (long)argv[j]->ptr);
e2641e09	97	} else {
	98	cmdrepr = sdscatrepr(cmdrepr,(char*)argv[j]->ptr,
	99	sdslen(argv[j]->ptr));
	100	}
	101	if (j != argc-1)
	102	cmdrepr = sdscatlen(cmdrepr," ",1);
	103	}
	104	cmdrepr = sdscatlen(cmdrepr,"\r\n",2);
	105	cmdobj = createObject(REDIS_STRING,cmdrepr);
	106
	107	listRewind(monitors,&li);
	108	while((ln = listNext(&li))) {
	109	redisClient *monitor = ln->value;
	110	addReply(monitor,cmdobj);
	111	}
	112	decrRefCount(cmdobj);
	113	}
	114
e2641e09	115	void syncCommand(redisClient *c) {
	116	/* ignore SYNC if aleady slave or in monitor mode */
	117	if (c->flags & REDIS_SLAVE) return;
	118
778b2210	119	/* Refuse SYNC requests if we are a slave but the link with our master
	120	* is not ok... */
	121	if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED) {
3ab20376	122	addReplyError(c,"Can't SYNC while not connected with my master");
778b2210	123	return;
	124	}
	125
e2641e09	126	/* SYNC can't be issued when the server has pending data to send to
	127	* the client about already issued commands. We need a fresh reply
	128	* buffer registering the differences between the BGSAVE and the current
	129	* dataset, so that we can copy to other slaves if needed. */
	130	if (listLength(c->reply) != 0) {
3ab20376	131	addReplyError(c,"SYNC is invalid with pending input");
e2641e09	132	return;
	133	}
	134
	135	redisLog(REDIS_NOTICE,"Slave ask for synchronization");
	136	/* Here we need to check if there is a background saving operation
	137	* in progress, or if it is required to start one */
	138	if (server.bgsavechildpid != -1) {
	139	/* Ok a background save is in progress. Let's check if it is a good
	140	* one for replication, i.e. if there is another slave that is
	141	* registering differences since the server forked to save */
	142	redisClient *slave;
	143	listNode *ln;
	144	listIter li;
	145
	146	listRewind(server.slaves,&li);
	147	while((ln = listNext(&li))) {
	148	slave = ln->value;
	149	if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) break;
	150	}
	151	if (ln) {
	152	/* Perfect, the server is already registering differences for
	153	* another slave. Set the right state, and copy the buffer. */
	154	listRelease(c->reply);
	155	c->reply = listDup(slave->reply);
	156	c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
	157	redisLog(REDIS_NOTICE,"Waiting for end of BGSAVE for SYNC");
	158	} else {
	159	/* No way, we need to wait for the next BGSAVE in order to
	160	* register differences */
	161	c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
	162	redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
	163	}
	164	} else {
	165	/* Ok we don't have a BGSAVE in progress, let's start one */
	166	redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC");
	167	if (rdbSaveBackground(server.dbfilename) != REDIS_OK) {
	168	redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
3ab20376	169	addReplyError(c,"Unable to perform background save");
e2641e09	170	return;
	171	}
	172	c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
	173	}
	174	c->repldbfd = -1;
	175	c->flags \|= REDIS_SLAVE;
	176	c->slaveseldb = 0;
	177	listAddNodeTail(server.slaves,c);
	178	return;
	179	}
	180
	181	void sendBulkToSlave(aeEventLoop el, int fd, void privdata, int mask) {
	182	redisClient *slave = privdata;
	183	REDIS_NOTUSED(el);
	184	REDIS_NOTUSED(mask);
	185	char buf[REDIS_IOBUF_LEN];
	186	ssize_t nwritten, buflen;
	187
	188	if (slave->repldboff == 0) {
	189	/* Write the bulk write count before to transfer the DB. In theory here
	190	* we don't know how much room there is in the output buffer of the
	191	* socket, but in pratice SO_SNDLOWAT (the minimum count for output
	192	* operations) will never be smaller than the few bytes we need. */
	193	sds bulkcount;
	194
	195	bulkcount = sdscatprintf(sdsempty(),"$%lld\r\n",(unsigned long long)
	196	slave->repldbsize);
	197	if (write(fd,bulkcount,sdslen(bulkcount)) != (signed)sdslen(bulkcount))
	198	{
	199	sdsfree(bulkcount);
	200	freeClient(slave);
	201	return;
	202	}
	203	sdsfree(bulkcount);
	204	}
	205	lseek(slave->repldbfd,slave->repldboff,SEEK_SET);
	206	buflen = read(slave->repldbfd,buf,REDIS_IOBUF_LEN);
	207	if (buflen <= 0) {
	208	redisLog(REDIS_WARNING,"Read error sending DB to slave: %s",
	209	(buflen == 0) ? "premature EOF" : strerror(errno));
	210	freeClient(slave);
	211	return;
	212	}
	213	if ((nwritten = write(fd,buf,buflen)) == -1) {
	214	redisLog(REDIS_VERBOSE,"Write error sending DB to slave: %s",
	215	strerror(errno));
	216	freeClient(slave);
	217	return;
	218	}
	219	slave->repldboff += nwritten;
	220	if (slave->repldboff == slave->repldbsize) {
	221	close(slave->repldbfd);
	222	slave->repldbfd = -1;
	223	aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
	224	slave->replstate = REDIS_REPL_ONLINE;
	225	if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
	226	sendReplyToClient, slave) == AE_ERR) {
	227	freeClient(slave);
	228	return;
	229	}
	230	addReplySds(slave,sdsempty());
	231	redisLog(REDIS_NOTICE,"Synchronization with slave succeeded");
	232	}
	233	}
234
235	/* This function is called at the end of every backgrond saving.
236	* The argument bgsaveerr is REDIS_OK if the background saving succeeded
237	* otherwise REDIS_ERR is passed to the function.
238	*
239	* The goal of this function is to handle slaves waiting for a successful
240	* background saving in order to perform non-blocking synchronization. */
241	void updateSlavesWaitingBgsave(int bgsaveerr) {
242	listNode *ln;
243	int startbgsave = 0;
244	listIter li;
245
246	listRewind(server.slaves,&li);
247	while((ln = listNext(&li))) {
248	redisClient *slave = ln->value;
249
250	if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) {
251	startbgsave = 1;
252	slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
253	} else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
254	struct redis_stat buf;
255
256	if (bgsaveerr != REDIS_OK) {
257	freeClient(slave);
258	redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error");
259	continue;
260	}
261	if ((slave->repldbfd = open(server.dbfilename,O_RDONLY)) == -1 \|\|
262	redis_fstat(slave->repldbfd,&buf) == -1) {
263	freeClient(slave);
264	redisLog(REDIS_WARNING,"SYNC failed. Can't open/stat DB after BGSAVE: %s", strerror(errno));
265	continue;
266	}
267	slave->repldboff = 0;
268	slave->repldbsize = buf.st_size;
269	slave->replstate = REDIS_REPL_SEND_BULK;
270	aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
271	if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
272	freeClient(slave);
273	continue;
274	}
275	}
276	}
277	if (startbgsave) {
278	if (rdbSaveBackground(server.dbfilename) != REDIS_OK) {
279	listIter li;
280
281	listRewind(server.slaves,&li);
282	redisLog(REDIS_WARNING,"SYNC failed. BGSAVE failed");
283	while((ln = listNext(&li))) {
284	redisClient *slave = ln->value;
285
286	if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START)
287	freeClient(slave);
288	}
289	}
290	}
291	}
292
f4aa600b	293	/* ----------------------------------- SLAVE -------------------------------- */
	294
	295	/* Abort the async download of the bulk dataset while SYNC-ing with master */
	296	void replicationAbortSyncTransfer(void) {
	297	redisAssert(server.replstate == REDIS_REPL_TRANSFER);
	298
	299	aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
	300	close(server.repl_transfer_s);
	301	close(server.repl_transfer_fd);
	302	unlink(server.repl_transfer_tmpfile);
	303	zfree(server.repl_transfer_tmpfile);
	304	server.replstate = REDIS_REPL_CONNECT;
	305	}
	306
	307	/* Asynchronously read the SYNC payload we receive from a master */
	308	void readSyncBulkPayload(aeEventLoop el, int fd, void privdata, int mask) {
26b33669	309	char buf[4096];
62ec599c	310	ssize_t nread, readlen;
	311	REDIS_NOTUSED(el);
	312	REDIS_NOTUSED(privdata);
	313	REDIS_NOTUSED(mask);
f4aa600b	314
26b33669	315	/* If repl_transfer_left == -1 we still have to read the bulk length
	316	* from the master reply. */
	317	if (server.repl_transfer_left == -1) {
	318	if (syncReadLine(fd,buf,1024,3600) == -1) {
	319	redisLog(REDIS_WARNING,
	320	"I/O error reading bulk count from MASTER: %s",
	321	strerror(errno));
	322	replicationAbortSyncTransfer();
	323	return;
	324	}
	325	if (buf[0] == '-') {
	326	redisLog(REDIS_WARNING,
	327	"MASTER aborted replication with an error: %s",
	328	buf+1);
	329	replicationAbortSyncTransfer();
	330	return;
89a1433e	331	} else if (buf[0] == '\0') {
	332	/* At this stage just a newline works as a PING in order to take
	333	* the connection live. So we refresh our last interaction
	334	* timestamp. */
	335	server.repl_transfer_lastio = time(NULL);
	336	return;
26b33669	337	} else if (buf[0] != '$') {
	338	redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$', are you sure the host and port are right?");
	339	replicationAbortSyncTransfer();
	340	return;
	341	}
	342	server.repl_transfer_left = strtol(buf+1,NULL,10);
	343	redisLog(REDIS_NOTICE,
	344	"MASTER <-> SLAVE sync: receiving %ld bytes from master",
	345	server.repl_transfer_left);
	346	return;
	347	}
	348
	349	/* Read bulk data */
62ec599c	350	readlen = (server.repl_transfer_left < (signed)sizeof(buf)) ?
62ec599c	351	server.repl_transfer_left : (signed)sizeof(buf);
f4aa600b	352	nread = read(fd,buf,readlen);
	353	if (nread <= 0) {
	354	redisLog(REDIS_WARNING,"I/O error trying to sync with MASTER: %s",
	355	(nread == -1) ? strerror(errno) : "connection lost");
	356	replicationAbortSyncTransfer();
	357	return;
	358	}
	359	server.repl_transfer_lastio = time(NULL);
	360	if (write(server.repl_transfer_fd,buf,nread) != nread) {
	361	redisLog(REDIS_WARNING,"Write error or short write writing to the DB dump file needed for MASTER <-> SLAVE synchrnonization: %s", strerror(errno));
	362	replicationAbortSyncTransfer();
	363	return;
	364	}
	365	server.repl_transfer_left -= nread;
	366	/* Check if the transfer is now complete */
	367	if (server.repl_transfer_left == 0) {
	368	if (rename(server.repl_transfer_tmpfile,server.dbfilename) == -1) {
	369	redisLog(REDIS_WARNING,"Failed trying to rename the temp DB into dump.rdb in MASTER <-> SLAVE synchronization: %s", strerror(errno));
	370	replicationAbortSyncTransfer();
	371	return;
	372	}
f6433915	373	redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Loading DB in memory");
f4aa600b	374	emptyDb();
9fd01051	375	/* Before loading the DB into memory we need to delete the readable
	376	* handler, otherwise it will get called recursively since
	377	* rdbLoad() will call the event loop to process events from time to
	378	* time for non blocking loading. */
	379	aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
f4aa600b	380	if (rdbLoad(server.dbfilename) != REDIS_OK) {
	381	redisLog(REDIS_WARNING,"Failed trying to load the MASTER synchronization DB from disk");
	382	replicationAbortSyncTransfer();
	383	return;
	384	}
	385	/* Final setup of the connected slave <- master link */
f4aa600b	386	zfree(server.repl_transfer_tmpfile);
	387	close(server.repl_transfer_fd);
	388	server.master = createClient(server.repl_transfer_s);
	389	server.master->flags \|= REDIS_MASTER;
	390	server.master->authenticated = 1;
	391	server.replstate = REDIS_REPL_CONNECTED;
26b33669	392	redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Finished with success");
f4aa600b	393	}
	394	}
	395
a3309139 PN	396	void syncWithMaster(aeEventLoop el, int fd, void privdata, int mask) {
a3309139 PN	397	char buf[1024], tmpfile[256];
e2641e09	398	int dfd, maxtries = 5;
a3309139 PN	399	REDIS_NOTUSED(el);
	400	REDIS_NOTUSED(privdata);
	401	REDIS_NOTUSED(mask);
e2641e09	402
a3309139 PN	403	/* Remove this event before anything else. */
a3309139 PN	404	aeDeleteFileEvent(server.el,fd,AE_READABLE\|AE_WRITABLE);
e2641e09	405
	406	/* AUTH with the master if required. */
	407	if(server.masterauth) {
a3309139 PN	408	char authcmd[1024];
	409	size_t authlen;
	410
	411	authlen = snprintf(authcmd,sizeof(authcmd),"AUTH %s\r\n",server.masterauth);
	412	if (syncWrite(fd,authcmd,authlen,5) == -1) {
e2641e09	413	redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s",
e2641e09	414	strerror(errno));
a3309139 PN	415	goto error;
a3309139 PN	416	}
e2641e09	417	/* Read the AUTH result. */
e2641e09	418	if (syncReadLine(fd,buf,1024,3600) == -1) {
e2641e09	419	redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s",
e2641e09	420	strerror(errno));
a3309139	421	goto error;
e2641e09	422	}
e2641e09	423	if (buf[0] != '+') {
e2641e09	424	redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?");
a3309139	425	goto error;
e2641e09	426	}
	427	}
	428
	429	/* Issue the SYNC command */
	430	if (syncWrite(fd,"SYNC \r\n",7,5) == -1) {
e2641e09	431	redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s",
e2641e09	432	strerror(errno));
a3309139	433	goto error;
e2641e09	434	}
26b33669	435
26b33669	436	/* Prepare a suitable temp file for bulk transfer */
e2641e09	437	while(maxtries--) {
	438	snprintf(tmpfile,256,
	439	"temp-%d.%ld.rdb",(int)time(NULL),(long int)getpid());
	440	dfd = open(tmpfile,O_CREAT\|O_WRONLY\|O_EXCL,0644);
	441	if (dfd != -1) break;
	442	sleep(1);
	443	}
	444	if (dfd == -1) {
e2641e09	445	redisLog(REDIS_WARNING,"Opening the temp file needed for MASTER <-> SLAVE synchronization: %s",strerror(errno));
a3309139	446	goto error;
e2641e09	447	}
e2641e09	448
f4aa600b	449	/* Setup the non blocking download of the bulk file. */
a3309139	450	if (aeCreateFileEvent(server.el,fd, AE_READABLE,readSyncBulkPayload,NULL)
62ec599c	451	== AE_ERR)
f4aa600b	452	{
f4aa600b	453	redisLog(REDIS_WARNING,"Can't create readable event for SYNC");
a3309139	454	goto error;
e2641e09	455	}
a3309139	456
f4aa600b	457	server.replstate = REDIS_REPL_TRANSFER;
26b33669	458	server.repl_transfer_left = -1;
f4aa600b	459	server.repl_transfer_fd = dfd;
	460	server.repl_transfer_lastio = time(NULL);
	461	server.repl_transfer_tmpfile = zstrdup(tmpfile);
a3309139 PN	462	return;
	463
	464	error:
	465	server.replstate = REDIS_REPL_CONNECT;
	466	close(fd);
	467	return;
	468	}
	469
	470	int connectWithMaster(void) {
	471	int fd;
	472
	473	fd = anetTcpNonBlockConnect(NULL,server.masterhost,server.masterport);
	474	if (fd == -1) {
	475	redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s",
	476	strerror(errno));
	477	return REDIS_ERR;
	478	}
	479
	480	if (aeCreateFileEvent(server.el,fd,AE_READABLE\|AE_WRITABLE,
	481	syncWithMaster,NULL) == AE_ERR)
	482	{
	483	close(fd);
	484	redisLog(REDIS_WARNING,"Can't create readable event for SYNC");
	485	return REDIS_ERR;
	486	}
	487
	488	server.repl_transfer_s = fd;
	489	server.replstate = REDIS_REPL_CONNECTING;
e2641e09	490	return REDIS_OK;
	491	}
	492
	493	void slaveofCommand(redisClient *c) {
	494	if (!strcasecmp(c->argv[1]->ptr,"no") &&
	495	!strcasecmp(c->argv[2]->ptr,"one")) {
	496	if (server.masterhost) {
	497	sdsfree(server.masterhost);
	498	server.masterhost = NULL;
	499	if (server.master) freeClient(server.master);
f4aa600b	500	if (server.replstate == REDIS_REPL_TRANSFER)
f4aa600b	501	replicationAbortSyncTransfer();
e2641e09	502	server.replstate = REDIS_REPL_NONE;
	503	redisLog(REDIS_NOTICE,"MASTER MODE enabled (user request)");
	504	}
	505	} else {
	506	sdsfree(server.masterhost);
	507	server.masterhost = sdsdup(c->argv[1]->ptr);
	508	server.masterport = atoi(c->argv[2]->ptr);
	509	if (server.master) freeClient(server.master);
f4aa600b	510	if (server.replstate == REDIS_REPL_TRANSFER)
f4aa600b	511	replicationAbortSyncTransfer();
e2641e09	512	server.replstate = REDIS_REPL_CONNECT;
	513	redisLog(REDIS_NOTICE,"SLAVE OF %s:%d enabled (user request)",
	514	server.masterhost, server.masterport);
	515	}
	516	addReply(c,shared.ok);
	517	}
f4aa600b	518
	519	/* --------------------------- REPLICATION CRON ---------------------------- */
	520
89a1433e	521	#define REDIS_REPL_TIMEOUT 60
89a1433e	522	#define REDIS_REPL_PING_SLAVE_PERIOD 10
f4aa600b	523
	524	void replicationCron(void) {
	525	/* Bulk transfer I/O timeout? */
	526	if (server.masterhost && server.replstate == REDIS_REPL_TRANSFER &&
89a1433e	527	(time(NULL)-server.repl_transfer_lastio) > REDIS_REPL_TIMEOUT)
f4aa600b	528	{
	529	redisLog(REDIS_WARNING,"Timeout receiving bulk data from MASTER...");
	530	replicationAbortSyncTransfer();
	531	}
	532
89a1433e	533	/* Timed out master when we are an already connected slave? */
	534	if (server.masterhost && server.replstate == REDIS_REPL_CONNECTED &&
	535	(time(NULL)-server.master->lastinteraction) > REDIS_REPL_TIMEOUT)
	536	{
	537	redisLog(REDIS_WARNING,"MASTER time out: no data nor PING received...");
	538	freeClient(server.master);
	539	}
	540
f4aa600b	541	/* Check if we should connect to a MASTER */
62ec599c	542	if (server.replstate == REDIS_REPL_CONNECT) {
f4aa600b	543	redisLog(REDIS_NOTICE,"Connecting to MASTER...");
a3309139 PN	544	if (connectWithMaster() == REDIS_OK) {
a3309139 PN	545	redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started");
f4aa600b	546	if (server.appendonly) rewriteAppendOnlyFileBackground();
	547	}
	548	}
89a1433e	549
	550	/* If we have attached slaves, PING them from time to time.
	551	* So slaves can implement an explicit timeout to masters, and will
	552	* be able to detect a link disconnection even if the TCP connection
	553	* will not actually go down. */
	554	if (!(server.cronloops % (REDIS_REPL_PING_SLAVE_PERIOD*10))) {
	555	listIter li;
	556	listNode *ln;
	557
	558	listRewind(server.slaves,&li);
	559	while((ln = listNext(&li))) {
	560	redisClient *slave = ln->value;
	561
	562	/* Don't ping slaves that are in the middle of a bulk transfer
	563	* with the master for first synchronization. */
	564	if (slave->replstate == REDIS_REPL_SEND_BULK) continue;
	565	if (slave->replstate == REDIS_REPL_ONLINE) {
	566	/* If the slave is online send a normal ping */
	567	addReplySds(slave,sdsnew("PING\r\n"));
	568	} else {
	569	/* Otherwise we are in the pre-synchronization stage.
	570	* Just a newline will do the work of refreshing the
	571	* connection last interaction time, and at the same time
	572	* we'll be sure that being a single char there are no
	573	* short-write problems. */
f96a9f82	574	if (write(slave->fd, "\n", 1) == -1) {
	575	/* Don't worry, it's just a ping. */
	576	}
89a1433e	577	}
	578	}
	579	}
f4aa600b	580	}