Redis Sentinel源码分析（二） - c++编程基础

__:hello”频道 if ((ri->flags & (SRI_MASTER|SRI_SLAVE)) && ri->pc == NULL) { ...... retval = redisAsyncCommand(ri->pc, sentinelReceiveHelloMessages, NULL, "SUBSCRIBE %s", SENTINEL_HELLO_CHANNEL); ...... } ...... }
sentinelPingInstance会根据instance状况，向其发送命令，可能是INFO/PING/PUBLISH

void sentinelPingInstance(sentinelRedisInstance *ri) {
    //假如instance处于不可连接状态或者过多的命令（100）还没有发送出去，直接返回
    if (ri->flags & SRI_DISCONNECTED) return;
    if (ri->pending_commands >= SENTINEL_MAX_PENDING_COMMANDS) return;
    //对于slave instance，如果其master处于异常状态（SRI_O_DOWN|SRI_FAILOVER_IN_PROGRESS），则向该slave发送info的频率从10s一发提高到1s一发
    if ((ri->flags & SRI_SLAVE) &&
        (ri->master->flags & (SRI_O_DOWN|SRI_FAILOVER_IN_PROGRESS))) {
        info_period = 1000;
    } else {
        info_period = SENTINEL_INFO_PERIOD;
    }
    //对于mastere/slave instance，每隔info_period时间，向其发送info命令，注册info命令的回调函数为sentinelInfoReplyCallback
    //sentinelInfoReplyCallback会根据从master/slave所得到的回复中分析出相关信息，并更新sentinelRedisInstance的当前状态
    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) return;
        ri->pending_commands++;
    } 
    //对于所有类型的instance，都定时向其发送PING命令（1s）,注册ping命令的回调函数为sentinelPingReplyCallback
    //sentinelPingReplyCallback根据PING命令的返回值判断instance当前状态
	else if ((now - ri->last_pong_time) > SENTINEL_PING_PERIOD) {
        retval = redisAsyncCommand(ri->cc,
            sentinelPingReplyCallback, NULL, "PING");
        if (retval != REDIS_OK) return;
        ri->pending_commands++;
    //每隔2s向master/slave的“__sentinel__:hello”频道发布消息
    //消息内容为：ip,port,runid,current_epoch, master->name,master->ip,master->port
    } else if ((ri->flags & SRI_SENTINEL) == 0 &&
               (now - ri->last_pub_time) > SENTINEL_PUBLISH_PERIOD)
    {
        sentinelSendHello(ri);
    }
}

sentinelCheckObjectivelyDown函数确认是否将master状态从sdown改为odown

void sentinelCheckObjectivelyDown(sentinelRedisInstance *master) {
    ......
    //假如本身的状态为sdown，则开始判断是否可以判断为odown
    if (master->flags & SRI_S_DOWN) {
        quorum = 1; 
        di = dictGetIterator(master->sentinels);
        //遍历sentinel字典，查看其是否将master状态职位sdown
        while((de = dictNext(di)) != NULL) {
            sentinelRedisInstance *ri = dictGetVal(de);
            if (ri->flags & SRI_MASTER_DOWN) quorum++;
        }
        dictReleaseIterator(di);
        //假如sentinel flag状态为SRI_MASTER_DOWN的sentinel个数达到用户定义的quorum个数，则将master状态置为odown
        if (quorum >= master->quorum) odown = 1;
    }
	......
}

sentinelStartFailoverIfNeeded函数heck是否需要做fail over，如果确认需要，则调用sentinelStartFailover修改自身状态

int sentinelStartFailoverIfNeeded(sentinelRedisInstance *master) {
    //确认master状态为odown
    if (!(master->flags & SRI_O_DOWN)) return 0;
    //确认failover没有在运行
    if (master->flags & SRI_FAILOVER_IN_PROGRESS) return 0;
    //确认在超时时间*2内没有failover在运行
    if (mstime() - master->failover_start_time <
        master->failover_timeout*2) return 0;
    sentinelStartFailover(master);
    return 1;
}

在确认要进行failover后，调用sentinelStartFailover修改相关状态数据

void sentinelStartFailover(sentinelRedisInstance *master) {
    redisAssert(master->flags & SRI_MASTER);
    // 设置 failover 状态
    master->failover_state = SENTINEL_FAILOVER_STATE_WAIT_START;
    // 设置master当前状态
    master->flags |= SRI_FAILOVER_IN_PROGRESS;
    // 设置failover_epoch
    master->failover_epoch = ++sentinel.c

Redis Sentinel源码分析（二）(二)