Redis源码解析:21sentinel(二)定期发送消息、检测主观下线

六:定时发送消息

         哨兵每隔一段时间,会向其所监控的所有实例发送一些命令,用于获取这些实例的状态。这些命令包括:”PING”、”INFO”和”PUBLISH”。

         “PING”命令,主要用于哨兵探测实例是否活着。如果对方超过一段时间,还没有回复”PING”命令,则认为其是主观下线了。

         “INFO”命令,主要用于哨兵获取实例当前的状态和信息,比如该实例当前是主节点还是从节点;该实例反馈的IP地址和PORT信息,是否与我记录的一样;该实例如果是主节点的话,那它都有哪些从节点;该实例如果是从节点的话,它与主节点是否连通,它的优先级是多少,它的复制偏移量是多少等等,这些信息在故障转移流程中,是判断实例状态的重要信息;

         “PUBLISH”命令,主要用于哨兵向实例的HELLO频道发布有关自己以及主节点的信息,也就是所谓的HELLO消息。因为所有哨兵都会订阅主节点和从节点的HELLO频道,因此,每个哨兵都会收到其他哨兵发布的信息。

         因此,通过这些命令,尽管在配置文件中只配置了主节点的信息,但是哨兵可以通过主节点的”INFO”回复,得到所有从节点的信息;又可以通过订阅实例的HELLO频道,接收其他哨兵通过”PUBLISH”命令发布的信息,从而得到监控同一主节点的所有其他哨兵的信息。

 

         在“主函数”sentinelHandleRedisInstance中,是通过调用sentinelSendPeriodicCommands来发送这些命令的。注意,以上的命令都有自己的发送周期,在sentinelSendPeriodicCommands函数中,并不是一并发送三个命令,而是发送那些,按照发送周期应该发送的命令。该函数的代码如下:

void sentinelSendPeriodicCommands(sentinelRedisInstance *ri) {
    mstime_t now = mstime();
    mstime_t info_period, ping_period;
    int retval;

    /* Return ASAP if we have already a PING or INFO already pending, or
     * in the case the instance is not properly connected. */
    if (ri->flags & SRI_DISCONNECTED) return;

    /* For INFO, PING, PUBLISH that are not critical commands to send we
     * also have a limit of SENTINEL_MAX_PENDING_COMMANDS. We don't
     * want to use a lot of memory just because a link is not working
     * properly (note that anyway there is a redundant protection about this,
     * that is, the link will be disconnected and reconnected if a long
     * timeout condition is detected. */
    if (ri->pending_commands >= SENTINEL_MAX_PENDING_COMMANDS) return;

    /* If this is a slave of a master in O_DOWN condition we start sending
     * it INFO every second, instead of the usual SENTINEL_INFO_PERIOD
     * period. In this state we want to closely monitor slaves in case they
     * are turned into masters by another Sentinel, or by the sysadmin. */
    if ((ri->flags & SRI_SLAVE) &&
        (ri->master->flags & (SRI_O_DOWN|SRI_FAILOVER_IN_PROGRESS))) {
        info_period = 1000;
    } else {
        info_period = SENTINEL_INFO_PERIOD;
    }

    /* We ping instances every time the last received pong is older than
     * the configured 'down-after-milliseconds' time, but every second
     * anyway if 'down-after-milliseconds' is greater than 1 second. */
    ping_period = ri->down_after_period;
    if (ping_period > SENTINEL_PING_PERIOD) ping_period = SENTINEL_PING_PERIOD;

    if ((ri->flags & SRI_SENTINEL) == 0 &&
        (ri->info_refresh == 0 ||
        (now - ri->info_refresh) > info_period))
    {
        /* Send INFO to masters and slaves, not sentinels. */
        retval = redisAsyncCommand(ri->cc,
            sentinelInfoReplyCallback, NULL, "INFO");
        if (retval == REDIS_OK) ri->pending_commands++;
    } else if ((now - ri->last_pong_time) > ping_period) {
        /* Send PING to all the three kinds of instances. */
        sentinelSendPing(ri);
    } else if ((now - ri->last_pub_time) > SENTINEL_PUBLISH_PERIOD) {
        /* PUBLISH hello messages to all the three kinds of instances. */
        sentinelSendHello(ri);
    }
}

         如果实例标志位中设置了SRI_DISCONNECTED标记,说明当前实例的异步上下文还没有创建好,因此直接返回;

         实例的pending_commands属性,表示已经向该实例发送的命令中,尚有pending_commands个命令还没有收到回复。每次调用redisAsyncCommand函数,向实例异步发送一条命令之后,就会增加该属性的值,而每当收到命令回复之后,就会减少该属性的值;

         因此,如果该属性的值大于SENTINEL_MAX_PENDING_COMMANDS(100),说明该实例尚有超过100条命令的回复信息没有收到。这种情况下,说明与实例的连接已经不正常了,为了节约内存,因此直接返回;

         接下来计算info_period和ping_period,这俩值表示发送"INFO"和"PING"命令的时间周期。如果当前时间距离上次收到"INFO"或"PING"回复的时间已经超过了info_period或ping_period,则向实例发送"INFO"或"PING"命令;

         如果当前实例为从节点,并且该从节点对应的主节点已经客观下线了,则置info_period为1000,否则的话置为SENTINEL_INFO_PERIOD(10000)。之所以在主节点客观下线后更频繁的向从节点发送"INFO"命令,是因为从节点可能会被置为新的主节点,因此需要更加实时的获取其状态;

         将ping_period置为ri->down_after_period的值,该属性的值是根据配置文件中down-after-milliseconds选项得到的,如果该属性值大于SENTINEL_PING_PERIOD(1000),则将ping_period置为SENTINEL_PING_PERIOD;

         接下来开始发送命令:如果当前实例不是哨兵实例,并且距离上次收到"INFO"命令回复已经超过了info_period,则向该实例异步发送"INFO"命令。

         否则,如果距离上次收到"PING"命令回复已经超过了ping_period,则调用函数sentinelSendPing向该实例异步发送"PING"命令;

         否则,如果距离上次收到"PUBLISH"命令的回复已经超过了SENTINEL_PUBLISH_PERIOD(2000),则调用函数sentinelSendHello向该实例异步发送"PUBLISH"命令;

         因此,"PING"用于探测实例是否活着,可以发送给所有类型的实例;而"INFO"命令用于获取实例的信息,只需发送给主节点和从节点实例;而"PUBLISH"用于向HELLO频道发布哨兵本身和主节点的信息,除了发送给主节点和从节点之外,哨兵本身也实现了"PUBLISH"命令的处理函数,因此该命令也会发送给哨兵实例。

 

1:PING消息

         函数sentinelSendPing用于向实例发送”PING”命令,因为该命令用于探测实例是否主观下线,因此等到后面讲解主观下线是在分析。

 

2:HELLO消息

         函数sentinelSendHello用于发布HELLO消息,它的代码如下:

int sentinelSendHello(sentinelRedisInstance *ri) {
    char ip[REDIS_IP_STR_LEN];
    char payload[REDIS_IP_STR_LEN+1024];
    int retval;
    char *announce_ip;
    int announce_port;
    sentinelRedisInstance *master = (ri->flags & SRI_MASTER) ? ri : ri->master;
    sentinelAddr *master_addr = sentinelGetCurrentMasterAddress(master);

    if (ri->flags & SRI_DISCONNECTED) return REDIS_ERR;

    /* Use the specified announce address if specified, otherwise try to
     * obtain our own IP address. */
    if (sentinel.announce_ip) {
        announce_ip = sentinel.announce_ip;
    } else {
        if (anetSockName(ri->cc->c.fd,ip,sizeof(ip),NULL) == -1)
            return REDIS_ERR;
        announce_ip = ip;
    }
    announce_port = sentinel.announce_port ?
                    sentinel.announce_port : server.port;

    /* Format and send the Hello message. */
    snprintf(payload,sizeof(payload),
        "%s,%d,%s,%llu," /* Info about this sentinel. */
        "%s,%s,%d,%llu", /* Info about current master. */
        announce_ip, announce_port, server.runid,
        (unsigned long long) sentinel.current_epoch,
        /* --- */
        master->name,master_addr->ip,master_addr->port,
        (unsigned long long) master->config_epoch);
    retval = redisAsyncCommand(ri->cc,
        sentinelPublishReplyCallback, NULL, "PUBLISH %s %s",
            SENTINEL_HELLO_CHANNEL,payload);
    if (retval != REDIS_OK) return REDIS_ERR;
    ri->pending_commands++;
    return REDIS_OK;
}

         首先得到实例ri所属的主节点实例master;然后调用sentinelGetCurrentMasterAddress函数得到master的地址信息;

         如果实例ri的标志位中具有SRI_DISCONNECTED标记的话,直接返回;

         如果当前哨兵配置了sentinel.announce_ip的话,则使用该ip信息作为自己的ip地址,否则,调用anetSockName函数,根据socket描述符得到当前哨兵的ip地址;

         如果当前哨兵配置了sentinel.announce_port的话,则使用该port信息作为自己的端口信息,否则,使用server.port作为当前哨兵的端口信息;

         接下来组装要发布的HELLO信息,HELLO信息的格式是:"sentinel_ip,sentinel_port,sentinel_runid,current_epoch,master_name,master_ip,master_port,master_config_epoch"

         接下来,向ri异步发送"PUBLISH__sentinel__:hello <HELLO>"命令,设置命令回调函数为sentinelPublishReplyCallback;

 

         当哨兵收到实例对于该”PUBLISH”命令的回复之后,会调用回调函数sentinelPublishReplyCallback,该函数只用于更新属性ri->last_pub_time,对回复内容无需关心:

void sentinelPublishReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
    sentinelRedisInstance *ri = c->data;
    redisReply *r;
    REDIS_NOTUSED(privdata);

    if (ri) ri->pending_commands--;
    if (!reply || !ri) return;
    r = reply;

    /* Only update pub_time if we actually published our message. Otherwise
     * we'll retry again in 100 milliseconds. */
    if (r->type != REDIS_REPLY_ERROR)
        ri->last_pub_time = mstime();
}

 

         之前在介绍sentinelReconnectInstance函数时讲过,当哨兵向主节点或从节点实例建立订阅连接时,向实例发送” SUBSCRIBE __sentinel__:hello"命令,订阅HELLO频道时,设置该命令的回调函数为sentinelReceiveHelloMessages。因此,当收到该频道上发布的消息时,就会调用函数sentinelReceiveHelloMessages。

         该频道上的消息,是监控同一实例的其他哨兵节点发来的HELLO消息,当前哨兵通过HELLO消息,来发现其他哨兵,并且相互之间交互最新的主节点信息。sentinelReceiveHelloMessages函数的代码如下:

void sentinelReceiveHelloMessages(redisAsyncContext *c, void *reply, void *privdata) {
    sentinelRedisInstance *ri = c->data;
    redisReply *r;
    REDIS_NOTUSED(privdata);

    if (!reply || !ri) return;
    r = reply;

    /* Update the last activity in the pubsub channel. Note that since we
     * receive our messages as well this timestamp can be used to detect
     * if the link is probably disconnected even if it seems otherwise. */
    ri->pc_last_activity = mstime();

    /* Sanity check in the reply we expect, so that the code that follows
     * can avoid to check for details. */
    if (r->type != REDIS_REPLY_ARRAY ||
        r->elements != 3 ||
        r->element[0]->type != REDIS_REPLY_STRING ||
        r->element[1]->type != REDIS_REPLY_STRING ||
        r->element[2]->type != REDIS_REPLY_STRING ||
        strcmp(r->element[0]->str,"message") != 0) return;

    /* We are not interested in meeting ourselves */
    if (strstr(r->element[2]->str,server.runid) != NULL) return;

    sentinelProcessHelloMessage(r->element[2]->str, r->element[2]->len);
}

         该函数中,首先更新ri->pc_last_activity为当前时间;

         然后判断是否处理接收到的消息,注意,只处理"message"消息,也就是说不会处理"subscribe"消息;

         注意,如果收到的"message"消息中,包含了自身的runid,说明这是本哨兵自己发送的消息,因此无需处理,直接返回;

         最后,调用sentinelProcessHelloMessage函数处理收到的HELLO消息;

         注意:在测试时发现会收到从节点重复的HELLO消息,也就是同一时间,同一个哨兵发布的两条一模一样的消息。这是因为哨兵向主节点发送的”PUBLISH”命令,会因为主从复制的原因,而同步到从节点;而同时该哨兵也向从节点发送”PUBLISH”命令,因此,从节点就会在同一时间,收到两条一模一样的HELLO消息,并将它们发布到频道上。

 

         另外,一旦哨兵发现了其他哨兵之后,可以直接向其发送"PUBLISH __sentinel__:hello <HELLO>"命令。哨兵自己实现了”PUBLISH”的处理函数sentinelPublishCommand,当收到其他哨兵直接发来的HELLO消息时,就会调用该函数处理。该函数的代码如下:

void sentinelPublishCommand(redisClient *c) {
    if (strcmp(c->argv[1]->ptr,SENTINEL_HELLO_CHANNEL)) {
        addReplyError(c, "Only HELLO messages are accepted by Sentinel instances.");
        return;
    }
    sentinelProcessHelloMessage(c->argv[2]->ptr,sdslen(c->argv[2]->ptr));
    addReplyLongLong(c,1);
}


         因此,不管是从真正的订阅频道中收到HELLO消息,还是直接收到其他哨兵发来的”PUBLISH”命令,最终都是通过sentinelProcessHelloMessage函数对HELLO消息进行处理的。该函数的代码如下:

void sentinelProcessHelloMessage(char *hello, int hello_len) {
    /* Format is composed of 8 tokens:
     * 0=ip,1=port,2=runid,3=current_epoch,4=master_name,
     * 5=master_ip,6=master_port,7=master_config_epoch. */
    int numtokens, port, removed, master_port;
    uint64_t current_epoch, master_config_epoch;
    char **token = sdssplitlen(hello, hello_len, ",", 1, &numtokens);
    sentinelRedisInstance *si, *master;

    if (numtokens == 8) {
        /* Obtain a reference to the master this hello message is about */
        master = sentinelGetMasterByName(token[4]);
        if (!master) goto cleanup; /* Unknown master, skip the message. */

        /* First, try to see if we already have this sentinel. */
        port = atoi(token[1]);
        master_port = atoi(token[6]);
        si = getSentinelRedisInstanceByAddrAndRunID(
                        master->sentinels,token[0],port,token[2]);
        current_epoch = strtoull(token[3],NULL,10);
        master_config_epoch = strtoull(token[7],NULL,10);

        if (!si) {
            /* If not, remove all the sentinels that have the same runid
             * OR the same ip/port, because it's either a restart or a
             * network topology change. */
            removed = removeMatchingSentinelsFromMaster(master,token[0],port,
                            token[2]);
            if (removed) {
                sentinelEvent(REDIS_NOTICE,"-dup-sentinel",master,
                    "%@ #duplicate of %s:%d or %s",
                    token[0],port,token[2]);
            }

            /* Add the new sentinel. */
            si = createSentinelRedisInstance(NULL,SRI_SENTINEL,
                            token[0],port,master->quorum,master);
            if (si) {
                sentinelEvent(REDIS_NOTICE,"+sentinel",si,"%@");
                /* The runid is NULL after a new instance creation and
                 * for Sentinels we don't have a later chance to fill it,
                 * so do it now. */
                si->runid = sdsnew(token[2]);
                sentinelFlushConfig();
            }
        }

        /* Update local current_epoch if received current_epoch is greater.*/
        if (current_epoch > sentinel.current_epoch) {
            sentinel.current_epoch = current_epoch;
            sentinelFlushConfig();
            sentinelEvent(REDIS_WARNING,"+new-epoch",master,"%llu",
                (unsigned long long) sentinel.current_epoch);
        }

        /* Update master info if received configuration is newer. */
        if (master->config_epoch < master_config_epoch) {
            master->config_epoch = master_config_epoch;
            if (master_port != master->addr->port ||
                strcmp(master->addr->ip, token[5]))
            {
                sentinelAddr *old_addr;

                sentinelEvent(REDIS_WARNING,"+config-update-from",si,"%@");
                sentinelEvent(REDIS_WARNING,"+switch-master",
                    master,"%s %s %d %s %d",
                    master->name,
                    master->addr->ip, master->addr->port,
                    token[5], master_port);

                old_addr = dupSentinelAddr(master->addr);
                sentinelResetMasterAndChangeAddress(master, token[5], master_port);
                sentinelCallClientReconfScript(master,
                    SENTINEL_OBSERVER,"start",
                    old_addr,master->addr);
                releaseSentinelAddr(old_addr);
            }
        }

        /* Update the state of the Sentinel. */
        if (si) si->last_hello_time = mstime();
    }

cleanup:
    sdsfreesplitres(token,numtokens);
}

         首先,根据消息中的master_name,调用函数sentinelGetMasterByName,在字典sentinel.masters中寻找相应的主节点实例master,如果找不到,则直接退出;

         然后,调用getSentinelRedisInstanceByAddrAndRunID函数,根据消息中的sentinel_ip,sentinel_port和sentinel_runid信息,在字典master->sentinels中,找到runid,ip和port都匹配的哨兵实例。

         如果没有找到匹配的哨兵实例,要么这是一个新发现的哨兵,要么是某个哨兵的信息发生了变化(比如有可能某个哨兵实例重启了,导致runid发生了变化;或者网络拓扑发生了变化,导致ip或port发生了变化)。

         这种情况下,首先调用函数removeMatchingSentinelsFromMaster,删除字典master->sentinels中,具有相同runid,或者具有相同ip和port的哨兵实例;然后根据HELLO消息中的ip和port信息,重新创建一个新的哨兵实例,添加到字典master->sentinels中,这样下次调用sentinelReconnectInstance时,就会向该哨兵实例进行建链了。;

         如果找到了匹配的哨兵实例,并且HELLO消息中的sentinel_current_epoch,大于本实例当前的current_epoch,则更新本实例的current_epoch属性;

         如果HELLO消息中的master_config_epoch,大于本实例记录的master的config_epoch,则更新本实例记录的master的config_epoch。并且如果HELLO消息中的master_ip或master_port,与本实例记录的主节点的ip或port信息不匹配的话,则说明可能发生了故障转移,某个从节点升级成为了新的主节点,因此调用sentinelResetMasterAndChangeAddress函数,重置主节点,及其从节点实例的信息;

         最后,更新si->last_hello_time属性为当前时间;

 

3:”INFO”命令

         “INFO”命令,主要用于哨兵获取主从节点实例当前的状态和信息,比如该实例当前是主节点还是从节点;该实例反馈的IP地址和PORT信息,是否与本哨兵记录的一样;该实例如果是主节点的话,那它都有哪些从节点;该实例如果是从节点的话,它与主节点是否连通,它的优先级是多少,它的复制偏移量是多少等等,这些信息在故障转移流程中,是判断实例状态的重要信息;

         在sentinelSendPeriodicCommands函数中,设置的”INFO”命令的回调函数是sentinelInfoReplyCallback。该函数的代码很简单,主要是调用sentinelRefreshInstanceInfo函数对回复进行处理。因此,主要看一下sentinelRefreshInstanceInfo函数的代码:

void sentinelRefreshInstanceInfo(sentinelRedisInstance *ri, const char *info) {
    sds *lines;
    int numlines, j;
    int role = 0;

    /* The following fields must be reset to a given value in the case they
     * are not found at all in the INFO output. */
    ri->master_link_down_time = 0;

    /* Process line by line. */
    lines = sdssplitlen(info,strlen(info),"\r\n",2,&numlines);
    for (j = 0; j < numlines; j++) {
        sentinelRedisInstance *slave;
        sds l = lines[j];

        /* run_id:<40 hex chars>*/
        if (sdslen(l) >= 47 && !memcmp(l,"run_id:",7)) {
            if (ri->runid == NULL) {
                ri->runid = sdsnewlen(l+7,40);
            } else {
                if (strncmp(ri->runid,l+7,40) != 0) {
                    sentinelEvent(REDIS_NOTICE,"+reboot",ri,"%@");
                    sdsfree(ri->runid);
                    ri->runid = sdsnewlen(l+7,40);
                }
            }
        }

        /* old versions: slave0:<ip>,<port>,<state>
         * new versions: slave0:ip=127.0.0.1,port=9999,... */
        if ((ri->flags & SRI_MASTER) &&
            sdslen(l) >= 7 &&
            !memcmp(l,"slave",5) && isdigit(l[5]))
        {
            char *ip, *port, *end;

            if (strstr(l,"ip=") == NULL) {
                /* Old format. */
                ip = strchr(l,':'); if (!ip) continue;
                ip++; /* Now ip points to start of ip address. */
                port = strchr(ip,','); if (!port) continue;
                *port = '\0'; /* nul term for easy access. */
                port++; /* Now port points to start of port number. */
                end = strchr(port,','); if (!end) continue;
                *end = '\0'; /* nul term for easy access. */
            } else {
                /* New format. */
                ip = strstr(l,"ip="); if (!ip) continue;
                ip += 3; /* Now ip points to start of ip address. */
                port = strstr(l,"port="); if (!port) continue;
                port += 5; /* Now port points to start of port number. */
                /* Nul term both fields for easy access. */
                end = strchr(ip,','); if (end) *end = '\0';
                end = strchr(port,','); if (end) *end = '\0';
            }

            /* Check if we already have this slave into our table,
             * otherwise add it. */
            if (sentinelRedisInstanceLookupSlave(ri,ip,atoi(port)) == NULL) {
                if ((slave = createSentinelRedisInstance(NULL,SRI_SLAVE,ip,
                            atoi(port), ri->quorum, ri)) != NULL)
                {
                    sentinelEvent(REDIS_NOTICE,"+slave",slave,"%@");
                    sentinelFlushConfig();
                }
            }
        }

        /* master_link_down_since_seconds:<seconds> */
        if (sdslen(l) >= 32 &&
            !memcmp(l,"master_link_down_since_seconds",30))
        {
            ri->master_link_down_time = strtoll(l+31,NULL,10)*1000;
        }

        /* role:<role> */
        if (!memcmp(l,"role:master",11)) role = SRI_MASTER;
        else if (!memcmp(l,"role:slave",10)) role = SRI_SLAVE;

        if (role == SRI_SLAVE) {
            /* master_host:<host> */
            if (sdslen(l) >= 12 && !memcmp(l,"master_host:",12)) {
                if (ri->slave_master_host == NULL ||
                    strcasecmp(l+12,ri->slave_master_host))
                {
                    sdsfree(ri->slave_master_host);
                    ri->slave_master_host = sdsnew(l+12);
                    ri->slave_conf_change_time = mstime();
                }
            }

            /* master_port:<port> */
            if (sdslen(l) >= 12 && !memcmp(l,"master_port:",12)) {
                int slave_master_port = atoi(l+12);

                if (ri->slave_master_port != slave_master_port) {
                    ri->slave_master_port = slave_master_port;
                    ri->slave_conf_change_time = mstime();
                }
            }

            /* master_link_status:<status> */
            if (sdslen(l) >= 19 && !memcmp(l,"master_link_status:",19)) {
                ri->slave_master_link_status =
                    (strcasecmp(l+19,"up") == 0) ?
                    SENTINEL_MASTER_LINK_STATUS_UP :
                    SENTINEL_MASTER_LINK_STATUS_DOWN;
            }

            /* slave_priority:<priority> */
            if (sdslen(l) >= 15 && !memcmp(l,"slave_priority:",15))
                ri->slave_priority = atoi(l+15);

            /* slave_repl_offset:<offset> */
            if (sdslen(l) >= 18 && !memcmp(l,"slave_repl_offset:",18))
                ri->slave_repl_offset = strtoull(l+18,NULL,10);
        }
    }
    ri->info_refresh = mstime();
    sdsfreesplitres(lines,numlines);

    /* ---------------------------- Acting half -----------------------------
     * Some things will not happen if sentinel.tilt is true, but some will
     * still be processed. */

    /* Remember when the role changed. */
    if (role != ri->role_reported) {
        ri->role_reported_time = mstime();
        ri->role_reported = role;
        if (role == SRI_SLAVE) ri->slave_conf_change_time = mstime();
        /* Log the event with +role-change if the new role is coherent or
         * with -role-change if there is a mismatch with the current config. */
        sentinelEvent(REDIS_VERBOSE,
            ((ri->flags & (SRI_MASTER|SRI_SLAVE)) == role) ?
            "+role-change" : "-role-change",
            ri, "%@ new reported role is %s",
            role == SRI_MASTER ? "master" : "slave",
            ri->flags & SRI_MASTER ? "master" : "slave");
    }

    /* None of the following conditions are processed when in tilt mode, so
     * return asap. */
    if (sentinel.tilt) return;

    /* Handle master -> slave role switch. */
    if ((ri->flags & SRI_MASTER) && role == SRI_SLAVE) {
        /* Nothing to do, but masters claiming to be slaves are
         * considered to be unreachable by Sentinel, so eventually
         * a failover will be triggered. */
    }
    ...
} 

         该函数首先在for循环中解析"INFO"回复信息:

         首先解析出"run_id"之后的信息,保存在ri->runid中。如果该实例的runid发生了变化,还需要记录日志,向"+reboot"频道发布消息;

         如果实例为主节点,则解析"slave"后的从节点信息,取出其中的ip和port信息,然后根据ip和port,调用sentinelRedisInstanceLookupSlave函数,在字典ri->slaves中寻找是否已经保存了该从节点的信息。如果没有,则调用createSentinelRedisInstance创建从节点实例,并插入到ri->slaves中,也就是发现了主节点属下的从节点,下次调用函数sentinelReconnectInstance时,就会向该从节点建链了;

         解析"master_link_down_since_seconds"信息,该信息表示从节点与主节点的断链时间。将其转换成整数后,记录到ri->master_link_down_time中;

         解析"role"信息,如果包含"role:master",则置role为SRI_MASTER,说明该实例报告自己为主节点;如果包含"role:slave",则置role为SRI_SLAVE,说明该实例报告自己为从节点;

         如果role为SRI_SLAVE,找到回复信息中的"master_host:"信息,记录到ri->slave_master_host中;找到回复信息中的"master_port:"信息,记录到ri->slave_master_port中;找到回复信息中的"master_link_status:"信息,根据其值是否为"up",记录到ri->slave_master_link_status中;找到回复信息中的"slave_priority:"信息,记录到ri->slave_priority中;找到回复信息中的"slave_repl_offset:"信息,记录到ri->slave_repl_offset中;

         解析完所有"INFO"回复信息之后,更新ri->info_refresh为当前时间;

        

         接下来根据实例的角色信息执行一些动作:

         ri->role_reported的初始值是根据ri->flags得到的,如果收到"INFO"回复后,解析得到的role与ri->role_reported不同,说明该实例的角色发生了变化,比如从主节点变成了从节点,或者相反。只要role与ri->role_reported不同,就首先更新ri->role_reported_time为当前时间,并且将ri->role_reported置为role;如果role为SRI_SLAVE,还需要更新ri->slave_conf_change_time的值为当前时间;最后,还根据ri->flags中的角色是否与role,来记录日志,发布信息;

         如果当前哨兵已经进入了TILT模式,则直接返回;

         如果ri->flags中为主节点,但是role为从节点,这种情况无需采取动作,因为这种情况会被视为主节点不可达,最终会引发故障迁移流程;

         本函数剩下的动作,与故障转移流程有关,后续在介绍。

 

七:判断实例是否主观下线

         首先解释一下主观下线和客观下线的区别。

         所谓主观下线,就是从“我”(当前实例)的角度来看,某个实例已经下线了。但是单个哨兵的视角可能是盲目的,仅从“我”的角度,就决定一个实例下线是武断的。因此,“我”还会通过命令询问其他哨兵节点,看它们是否也认为该实例已经下线了,如果超过quorum个(包括“我”)哨兵反馈认为该实例已经下线了,则“我”就会认为该实例确实已经下线了,也就是所谓的客观下线了。

 

         判断某个实例主观下线,主要是根据其是否能及时回复”PING”命令决定的。因此,首先看一下发送”PING”命令的函数sentinelSendPing的实现:

int sentinelSendPing(sentinelRedisInstance *ri) {
    int retval = redisAsyncCommand(ri->cc,
        sentinelPingReplyCallback, NULL, "PING");
    if (retval == REDIS_OK) {
        ri->pending_commands++;
        /* We update the ping time only if we received the pong for
         * the previous ping, otherwise we are technically waiting
         * since the first ping that did not received a reply. */
        if (ri->last_ping_time == 0) ri->last_ping_time = mstime();
        return 1;
    } else {
        return 0;
    }
}

         在该函数中,设置收到”PING”命令回复后的回调函数为sentinelPingReplyCallback。

需要注意的是,如果ri->last_ping_time值为0,则更新ri->last_ping_time为当前时间。而只有在收到"PING"命令的正常回复之后,ri->last_ping_time的值才会被置为0。

 

         下面是回调函数sentinelPingReplyCallback的代码:

void sentinelPingReplyCallback(redisAsyncContext *c, void *reply, void *privdata) {
    sentinelRedisInstance *ri = c->data;
    redisReply *r;
    REDIS_NOTUSED(privdata);

    if (ri) ri->pending_commands--;
    if (!reply || !ri) return;
    r = reply;

    if (r->type == REDIS_REPLY_STATUS ||
        r->type == REDIS_REPLY_ERROR) {
        /* Update the "instance available" field only if this is an
         * acceptable reply. */
        if (strncmp(r->str,"PONG",4) == 0 ||
            strncmp(r->str,"LOADING",7) == 0 ||
            strncmp(r->str,"MASTERDOWN",10) == 0)
        {
            ri->last_avail_time = mstime();
            ri->last_ping_time = 0; /* Flag the pong as received. */
        } else {
            /* Send a SCRIPT KILL command if the instance appears to be
             * down because of a busy script. */
            if (strncmp(r->str,"BUSY",4) == 0 &&
                (ri->flags & SRI_S_DOWN) &&
                !(ri->flags & SRI_SCRIPT_KILL_SENT))
            {
                if (redisAsyncCommand(ri->cc,
                        sentinelDiscardReplyCallback, NULL,
                        "SCRIPT KILL") == REDIS_OK)
                    ri->pending_commands++;
                ri->flags |= SRI_SCRIPT_KILL_SENT;
            }
        }
    }
    ri->last_pong_time = mstime();
}

         如果回复信息为"PONG","LOADING"或"MASTERDOWN",表示正常回复,因此置该实例的属性ri->last_avail_time为当前时间,并且置ri->last_ping_time为0,这样下次发送"PING"命令时就会更新ri->last_ping_time的值了;

         如果回复信息以"BUSY"开头,并且该实例已经被置为主观下线,并且还没有向该实例发送过"SCRIPT KILL"命令,则向该实例发送"SCRIPTKILL"命令;

         最后,不管回复信息是什么,更新ri->last_pong_time为当前时间。

 

         因此,有关”PING”命令的时间属性总结如下:

         ri->last_ping_time:上一次正常发送”PING”命令的时间。需要注意的是,只有当收到"PING"命令的正常回复后,下次发送"PING"命令时才会更新该属性为当时时间戳。如果发送”PING”命令后,没有收到任何回复,或者没有收到正常回复,则下次发送”PING”命令时,就不会更新该属性。如果该属性值为0,说明已经收到了上一个"PING"命令的正常回复,但是还没有开始发送下一个"PING"命令。检测实例是否主观下线,主要就是根据该属性判断的。

         ri->last_pong_time:每当收到"PING"命令的回复后,不管是否是正常恢复,都会更新该属性为当时时间戳;

 

         在哨兵的“主函数”sentinelHandleRedisInstance中,调用sentinelCheckSubjectivelyDown函数检测实例是否主观下线,该函数同时还会检测TCP连接是否正常。该函数的代码如下:

void sentinelCheckSubjectivelyDown(sentinelRedisInstance *ri) {
    mstime_t elapsed = 0;

    if (ri->last_ping_time)
        elapsed = mstime() - ri->last_ping_time;

    /* Check if we are in need for a reconnection of one of the
     * links, because we are detecting low activity.
     *
     * 1) Check if the command link seems connected, was connected not less
     *    than SENTINEL_MIN_LINK_RECONNECT_PERIOD, but still we have a
     *    pending ping for more than half the timeout. */
    if (ri->cc &&
        (mstime() - ri->cc_conn_time) > SENTINEL_MIN_LINK_RECONNECT_PERIOD &&
        ri->last_ping_time != 0 && /* Ther is a pending ping... */
        /* The pending ping is delayed, and we did not received
         * error replies as well. */
        (mstime() - ri->last_ping_time) > (ri->down_after_period/2) &&
        (mstime() - ri->last_pong_time) > (ri->down_after_period/2))
    {
        sentinelKillLink(ri,ri->cc);
    }

    /* 2) Check if the pubsub link seems connected, was connected not less
     *    than SENTINEL_MIN_LINK_RECONNECT_PERIOD, but still we have no
     *    activity in the Pub/Sub channel for more than
     *    SENTINEL_PUBLISH_PERIOD * 3.
     */
    if (ri->pc &&
        (mstime() - ri->pc_conn_time) > SENTINEL_MIN_LINK_RECONNECT_PERIOD &&
        (mstime() - ri->pc_last_activity) > (SENTINEL_PUBLISH_PERIOD*3))
    {
        sentinelKillLink(ri,ri->pc);
    }

    /* Update the SDOWN flag. We believe the instance is SDOWN if:
     *
     * 1) It is not replying.
     * 2) We believe it is a master, it reports to be a slave for enough time
     *    to meet the down_after_period, plus enough time to get two times
     *    INFO report from the instance. */
    if (elapsed > ri->down_after_period ||
        (ri->flags & SRI_MASTER &&
         ri->role_reported == SRI_SLAVE &&
         mstime() - ri->role_reported_time >
          (ri->down_after_period+SENTINEL_INFO_PERIOD*2)))
    {
        /* Is subjectively down */
        if ((ri->flags & SRI_S_DOWN) == 0) {
            sentinelEvent(REDIS_WARNING,"+sdown",ri,"%@");
            ri->s_down_since_time = mstime();
            ri->flags |= SRI_S_DOWN;
        }
    } else {
        /* Is subjectively up */
        if (ri->flags & SRI_S_DOWN) {
            sentinelEvent(REDIS_WARNING,"-sdown",ri,"%@");
            ri->flags &= ~(SRI_S_DOWN|SRI_SCRIPT_KILL_SENT);
        }
    }
}

         ri->cc_conn_time属性表示上一次向该实例发起命令类型的TCP建链的时间;ri->pc_conn_time属性表示上一次向该实例发起订阅类型的TCP建链的时间;

         首先计算elapsed的值,该值表示是当前时间与ri->last_ping_time之间的时间差;

         然后判断命令类型的TCP连接是否正常,不正常的条件是:距离上次建链时已经超过了SENTINEL_MIN_LINK_RECONNECT_PERIOD,并且上次发送"PING"后还没有收到正常回复,且当前时间与ri->last_ping_time之间的时间差已经超过了ri->down_after_period/2,并且距离上次收到任何"PING"回复的时间,已经超过了ri->down_after_period/2;

         如果命令类型的连接不正常了,则直接调用sentinelKillLink断开连接,释放异步上下文;

 

         然后判断订阅类型的TCP连接是否正常,不正常的条件是:距离上次建链时已经超过了SENTINEL_MIN_LINK_RECONNECT_PERIOD,并且距离上次收到订阅频道发来的任何消息的时间,已经超过了SENTINEL_PUBLISH_PERIOD*3;

         如果订阅类型的连接不正常了,则直接调用sentinelKillLink断开连接,释放异步上下文;

 

         如果elapsed的值大于ri->down_after_period,或者:当前实例我认为它是主节点,但是它的"INFO"回复中却报告自己是从节点,并且距离上次收到它在"INFO"回复中报告自己是从节点的时间,已经超过了ri->down_after_period+SENTINEL_INFO_PERIOD*2;

         满足以上任意一个条件,都认为该实例是主观下线了。因此:只要该实例还没有标志为主观下线,则将SRI_S_DOWN标记增加到实例标志位中,表示该实例主观下线;

         如果不满足以上条件,但是该实例之前已经被标记为主观下线了,则认为该实例主观上线了,去掉其标志位中的SRI_S_DOWN和SRI_SCRIPT_KILL_SENT标记;

posted @ 2016-05-29 13:05  gqtc  阅读(972)  评论(0编辑  收藏  举报