redis3.0源码解析之主从复制及哨兵机制
文章目录
- redis3.0源码解析之主从复制及哨兵机制
- 一、主从复制
- 二、哨兵机制
- 3.不解
一、主从复制
1:salveof ip port:客户端向从服务器发送salve命令,从服务器解析命令更新master的ip与端口
void slaveofCommand(redisClient *c) {
/* SLAVEOF is not allowed in cluster mode as replication is automatically
* configured using the current address of the master node. */
// 不允许在集群模式中使用
if (server.cluster_enabled) {
addReplyError(c,"SLAVEOF not allowed in cluster mode.");
return;
}
/* The special host/port combination "NO" "ONE" turns the instance
* into a master. Otherwise the new master address is set. */
// SLAVEOF NO ONE 让从服务器转为主服务器
if (!strcasecmp(c->argv[1]->ptr,"no") &&
!strcasecmp(c->argv[2]->ptr,"one")) {
if (server.masterhost) {
// 让服务器取消复制,成为主服务器
replicationUnsetMaster();
redisLog(REDIS_NOTICE,"MASTER MODE enabled (user request)");
}
} else {
long port;
// 获取端口参数
if ((getLongFromObjectOrReply(c, c->argv[2], &port, NULL) != REDIS_OK))
return;
/* Check if we are already attached to the specified slave */
// 检查输入的 host 和 port 是否服务器目前的主服务器
// 如果是的话,向客户端返回 +OK ,不做其他动作
if (server.masterhost && !strcasecmp(server.masterhost,c->argv[1]->ptr)
&& server.masterport == port) {
redisLog(REDIS_NOTICE,"SLAVE OF would result into synchronization with the master we are already connected with. No operation performed.");
addReplySds(c,sdsnew("+OK Already connected to specified master\r\n"));
return;
}
/* There was no previous master or the user specified a different one,
* we can continue. */
// 没有前任主服务器,或者客户端指定了新的主服务器
// 开始执行复制操作
replicationSetMaster(c->argv[1]->ptr, port);
redisLog(REDIS_NOTICE,"SLAVE OF %s:%d enabled (user request)",
server.masterhost, server.masterport);
}
addReply(c,shared.ok);
}
2:在定时器中从服务器向主服务器发起TCP连接
// 以非阻塞方式连接主服务器
int connectWithMaster(void) {
int fd;
// 连接主服务器
fd = anetTcpNonBlockConnect(NULL,server.masterhost,server.masterport);
if (fd == -1) {
redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s",
strerror(errno));
return REDIS_ERR;
}
// 监听主服务器 fd 的读和写事件,并绑定文件事件处理器
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;
}
// 初始化统计变量
server.repl_transfer_lastio = server.unixtime;
server.repl_transfer_s = fd;
// 将状态改为已连接
server.repl_state = REDIS_REPL_CONNECTING;
return REDIS_OK;
}
3:建立TCP连接之后从服务器同步向主服务器发送PING等待PONG;同步发送用户验证信息,以及当前使用的端口号。
4:从服务器向主服务器发送同步命令,并根据返回值确定从服务器是否需要接受主服务器的RDB文件
// 从服务器用于同步主服务器的回调函数
void syncWithMaster(aeEventLoop *el, int fd, void *privdata, int mask) {
char tmpfile[256], *err;
int dfd, maxtries = 5;
int sockerr = 0, psync_result;
socklen_t errlen = sizeof(sockerr);
REDIS_NOTUSED(el);
REDIS_NOTUSED(privdata);
REDIS_NOTUSED(mask);
/* If this event fired after the user turned the instance into a master
* with SLAVEOF NO ONE we must just return ASAP. */
// 如果处于 SLAVEOF NO ONE 模式,那么关闭 fd
if (server.repl_state == REDIS_REPL_NONE) {
close(fd);
return;
}
/* Check for errors in the socket. */
// 检查套接字错误
if (getsockopt(fd, SOL_SOCKET, SO_ERROR, &sockerr, &errlen) == -1)
sockerr = errno;
if (sockerr) {
aeDeleteFileEvent(server.el,fd,AE_READABLE|AE_WRITABLE);
redisLog(REDIS_WARNING,"Error condition on socket for SYNC: %s",
strerror(sockerr));
goto error;
}
/* If we were connecting, it's time to send a non blocking PING, we want to
* make sure the master is able to reply before going into the actual
* replication process where we have long timeouts in the order of
* seconds (in the meantime the slave would block). */
// 如果状态为 CONNECTING ,那么在进行初次同步之前,
// 向主服务器发送一个非阻塞的 PONG
// 因为接下来的 RDB 文件发送非常耗时,所以我们想确认主服务器真的能访问
if (server.repl_state == REDIS_REPL_CONNECTING) {
redisLog(REDIS_NOTICE,"Non blocking connect for SYNC fired the event.");
/* Delete the writable event so that the readable event remains
* registered and we can wait for the PONG reply. */
// 手动发送同步 PING ,暂时取消监听写事件
aeDeleteFileEvent(server.el,fd,AE_WRITABLE);
// 更新状态
server.repl_state = REDIS_REPL_RECEIVE_PONG;
/* Send the PING, don't check for errors at all, we have the timeout
* that will take care about this. */
// 同步发送 PING
syncWrite(fd,"PING\r\n",6,100);
// 返回,等待 PONG 到达
return;
}
/* Receive the PONG command. */
// 接收 PONG 命令
if (server.repl_state == REDIS_REPL_RECEIVE_PONG) {
char buf[1024];
/* Delete the readable event, we no longer need it now that there is
* the PING reply to read. */
// 手动同步接收 PONG ,暂时取消监听读事件
aeDeleteFileEvent(server.el,fd,AE_READABLE);
/* Read the reply with explicit timeout. */
// 尝试在指定时间限制内读取 PONG
buf[0] = '\0';
// 同步接收 PONG
if (syncReadLine(fd,buf,sizeof(buf),
server.repl_syncio_timeout*1000) == -1)
{
redisLog(REDIS_WARNING,
"I/O error reading PING reply from master: %s",
strerror(errno));
goto error;
}
/* We accept only two replies as valid, a positive +PONG reply
* (we just check for "+") or an authentication error.
* Note that older versions of Redis replied with "operation not
* permitted" instead of using a proper error code, so we test
* both. */
// 接收到的数据只有两种可能:
// 第一种是 +PONG ,第二种是因为未验证而出现的 -NOAUTH 错误
if (buf[0] != '+' &&
strncmp(buf,"-NOAUTH",7) != 0 &&
strncmp(buf,"-ERR operation not permitted",28) != 0)
{
// 接收到未验证错误
redisLog(REDIS_WARNING,"Error reply to PING from master: '%s'",buf);
goto error;
} else {
// 接收到 PONG
redisLog(REDIS_NOTICE,
"Master replied to PING, replication can continue...");
}
}
/* AUTH with the master if required. */
// 进行身份验证
if(server.masterauth) {
err = sendSynchronousCommand(fd,"AUTH",server.masterauth,NULL);
if (err[0] == '-') {
redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s",err);
sdsfree(err);
goto error;
}
sdsfree(err);
}
/* Set the slave port, so that Master's INFO command can list the
* slave listening port correctly. */
// 将从服务器的端口发送给主服务器,
// 使得主服务器的 INFO 命令可以显示从服务器正在监听的端口
{
sds port = sdsfromlonglong(server.port);
err = sendSynchronousCommand(fd,"REPLCONF","listening-port",port,
NULL);
sdsfree(port);
/* Ignore the error if any, not all the Redis versions support
* REPLCONF listening-port. */
if (err[0] == '-') {
redisLog(REDIS_NOTICE,"(Non critical) Master does not understand REPLCONF listening-port: %s", err);
}
sdsfree(err);
}
/* Try a partial resynchonization. If we don't have a cached master
* slaveTryPartialResynchronization() will at least try to use PSYNC
* to start a full resynchronization so that we get the master run id
* and the global offset, to try a partial resync at the next
* reconnection attempt. */
// 根据返回的结果决定是执行部分 resync ,还是 full-resync
psync_result = slaveTryPartialResynchronization(fd);
// 可以执行部分 resync
if (psync_result == PSYNC_CONTINUE) {
redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Master accepted a Partial Resynchronization.");
// 返回
return;
}
/* Fall back to SYNC if needed. Otherwise psync_result == PSYNC_FULLRESYNC
* and the server.repl_master_runid and repl_master_initial_offset are
* already populated. */
// 主服务器不支持 PSYNC ,发送 SYNC
if (psync_result == PSYNC_NOT_SUPPORTED) {
redisLog(REDIS_NOTICE,"Retrying with SYNC...");
// 向主服务器发送 SYNC 命令
if (syncWrite(fd,"SYNC\r\n",6,server.repl_syncio_timeout*1000) == -1) {
redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s",
strerror(errno));
goto error;
}
}
// 如果执行到这里,
// 那么 psync_result == PSYNC_FULLRESYNC 或 PSYNC_NOT_SUPPORTED
/* Prepare a suitable temp file for bulk transfer */
// 打开一个临时文件,用于写入和保存接下来从主服务器传来的 RDB 文件数据
while(maxtries--) {
snprintf(tmpfile,256,
"temp-%d.%ld.rdb",(int)server.unixtime,(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. */
// 设置一个读事件处理器,来读取主服务器的 RDB 文件
if (aeCreateFileEvent(server.el,fd, AE_READABLE,readSyncBulkPayload,NULL)
== AE_ERR)
{
redisLog(REDIS_WARNING,
"Can't create readable event for SYNC: %s (fd=%d)",
strerror(errno),fd);
goto error;
}
// 设置状态
server.repl_state = REDIS_REPL_TRANSFER;
// 更新统计信息
server.repl_transfer_size = -1;
server.repl_transfer_read = 0;
server.repl_transfer_last_fsync_off = 0;
server.repl_transfer_fd = dfd;
server.repl_transfer_lastio = server.unixtime;
server.repl_transfer_tmpfile = zstrdup(tmpfile);
return;
error:
close(fd);
server.repl_transfer_s = -1;
server.repl_state = REDIS_REPL_CONNECT;
return;
}
5:主服务器收到同步信息后根据id信息以及缓冲区是否发生覆盖来决定是否进行开启RDB,若开启RDB此时propagate传播到从服务器的信息会被暂存到客户端输出缓冲区中,等到RDB结束的时候重新触发客户端写事件将积压的缓冲区内容发送给从服务器,实现主从同步。
void syncCommand(redisClient *c) {
/* ignore SYNC if already slave or in monitor mode */
// 已经是 SLAVE ,或者处于 MONITOR 模式,返回
if (c->flags & REDIS_SLAVE) return;
/* Refuse SYNC requests if we are a slave but the link with our master
* is not ok... */
// 如果这是一个从服务器,但与主服务器的连接仍未就绪,那么拒绝 SYNC
if (server.masterhost && server.repl_state != 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. */
// 在客户端仍有输出数据等待输出,不能 SYNC
if (listLength(c->reply) != 0 || c->bufpos != 0) {
addReplyError(c,"SYNC and PSYNC are invalid with pending output");
return;
}
redisLog(REDIS_NOTICE,"Slave asks for synchronization");
/* Try a partial resynchronization if this is a PSYNC command.
* 如果这是一个 PSYNC 命令,那么尝试 partial resynchronization 。
*
* If it fails, we continue with usual full resynchronization, however
* when this happens masterTryPartialResynchronization() already
* replied with:
*
* 如果失败,那么使用 full resynchronization ,
* 在这种情况下, masterTryPartialResynchronization() 返回以下内容:
*
* +FULLRESYNC <runid> <offset>
*
* So the slave knows the new runid and offset to try a PSYNC later
* if the connection with the master is lost.
*
* 这样的话,之后如果主服务器断开,那么从服务器就可以尝试 PSYNC 了。
*/
if (!strcasecmp(c->argv[0]->ptr,"psync")) {
// 尝试进行 PSYNC
if (masterTryPartialResynchronization(c) == REDIS_OK) {
// 可执行 PSYNC
server.stat_sync_partial_ok++;
return; /* No full resync needed, return. */
} else {
// 不可执行 PSYNC
char *master_runid = c->argv[1]->ptr;
/* Increment stats for failed PSYNCs, but only if the
* runid is not "?", as this is used by slaves to force a full
* resync on purpose when they are not albe to partially
* resync. */
if (master_runid[0] != '?') server.stat_sync_partial_err++;
}
} else {
/* If a slave uses SYNC, we are dealing with an old implementation
* of the replication protocol (like redis-cli --slave). Flag the client
* so that we don't expect to receive REPLCONF ACK feedbacks. */
// 旧版实现,设置标识,避免接收 REPLCONF ACK
c->flags |= REDIS_PRE_PSYNC;
}
// 以下是完整重同步的情况。。。
/* Full resynchronization. */
// 执行 full resynchronization ,增加计数
server.stat_sync_full++;
/* Here we need to check if there is a background saving operation
* in progress, or if it is required to start one */
// 检查是否有 BGSAVE 在执行
if (server.rdb_child_pid != -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;
// 如果有至少一个 slave 在等待这个 BGSAVE 完成
// 那么说明正在进行的 BGSAVE 所产生的 RDB 也可以为其他 slave 所用
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. */
// 幸运的情况,可以使用目前 BGSAVE 所生成的 RDB
copyClientOutputBuffer(c,slave);
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 */
// 不好运的情况,必须等待下个 BGSAVE
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 */
// 没有 BGSAVE 在进行,开始一个新的 BGSAVE
redisLog(REDIS_NOTICE,"Starting BGSAVE for SYNC");
if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
redisLog(REDIS_NOTICE,"Replication failed, can't BGSAVE");
addReplyError(c,"Unable to perform background save");
return;
}
// 设置状态
c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
/* Flush the script cache for the new slave. */
// 因为新 slave 进入,刷新复制脚本缓存
replicationScriptCacheFlush();
}
if (server.repl_disable_tcp_nodelay)
anetDisableTcpNoDelay(NULL, c->fd); /* Non critical if it fails. */
c->repldbfd = -1;
c->flags |= REDIS_SLAVE;
server.slaveseldb = -1; /* Force to re-emit the SELECT command. */
// 添加到 slave 列表中
listAddNodeTail(server.slaves,c);
// 如果是第一个 slave ,那么初始化 backlog
if (listLength(server.slaves) == 1 && server.repl_backlog == NULL)
createReplicationBacklog();
return;
}
void backgroundSaveDoneHandler(int exitcode, int bysignal) {
// BGSAVE 成功
if (!bysignal && exitcode == 0) {
redisLog(REDIS_NOTICE,
"Background saving terminated with success");
server.dirty = server.dirty - server.dirty_before_bgsave;
server.lastsave = time(NULL);
server.lastbgsave_status = REDIS_OK;
// BGSAVE 出错
} else if (!bysignal && exitcode != 0) {
redisLog(REDIS_WARNING, "Background saving error");
server.lastbgsave_status = REDIS_ERR;
// BGSAVE 被中断
} else {
redisLog(REDIS_WARNING,
"Background saving terminated by signal %d", bysignal);
// 移除临时文件
rdbRemoveTempFile(server.rdb_child_pid);
/* SIGUSR1 is whitelisted, so we have a way to kill a child without
* tirggering an error conditon. */
if (bysignal != SIGUSR1)
server.lastbgsave_status = REDIS_ERR;
}
// 更新服务器状态
server.rdb_child_pid = -1;
server.rdb_save_time_last = time(NULL)-server.rdb_save_time_start;
server.rdb_save_time_start = -1;
/* Possibly there are slaves waiting for a BGSAVE in order to be served
* (the first stage of SYNC is a bulk transfer of dump.rdb) */
// 处理正在等待 BGSAVE 完成的那些 slave
updateSlavesWaitingBgsave(exitcode == 0 ? REDIS_OK : REDIS_ERR);
}
void updateSlavesWaitingBgsave(int bgsaveerr) {
listNode *ln;
int startbgsave = 0;
listIter li;
// 遍历所有 slave
listRewind(server.slaves,&li);
while((ln = listNext(&li))) {
redisClient *slave = ln->value;
if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) {
// 之前的 RDB 文件不能被 slave 使用,
// 开始新的 BGSAVE
startbgsave = 1;
slave->replstate = REDIS_REPL_WAIT_BGSAVE_END;
} else if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
// 执行到这里,说明有 slave 在等待 BGSAVE 完成
struct redis_stat buf;
// 但是 BGSAVE 执行错误
if (bgsaveerr != REDIS_OK) {
// 释放 slave
freeClient(slave);
redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error");
continue;
}
// 打开 RDB 文件
if ((slave->repldbfd = open(server.rdb_filename,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;
}
// 设置偏移量,各种值
slave->repldboff = 0;
slave->repldbsize = buf.st_size;
// 更新状态
slave->replstate = REDIS_REPL_SEND_BULK;
slave->replpreamble = sdscatprintf(sdsempty(),"$%lld\r\n",
(unsigned long long) slave->repldbsize);
// 清空之前的写事件处理器
aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
// 将 sendBulkToSlave 安装为 slave 的写事件处理器
// 它用于将 RDB 文件发送给 slave
if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE, sendBulkToSlave, slave) == AE_ERR) {
freeClient(slave);
continue;
}
}
}
// 需要执行新的 BGSAVE
if (startbgsave) {
/* Since we are starting a new background save for one or more slaves,
* we flush the Replication Script Cache to use EVAL to propagate every
* new EVALSHA for the first time, since all the new slaves don't know
* about previous scripts. */
// 开始行的 BGSAVE ,并清空脚本缓存
replicationScriptCacheFlush();
if (rdbSaveBackground(server.rdb_filename) != REDIS_OK) {
listIter li;
listRewind(server.slaves,&li);
redisLog(REDIS_WARNING,"SYNC failed. BGSAVE failed");
while((ln = listNext(&li))) {
redisClient *slave = ln->value;
if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START)
freeClient(slave);
}
}
}
}
void sendBulkToSlave(aeEventLoop *el, int fd, void *privdata, int mask) {
redisClient *slave = privdata;
REDIS_NOTUSED(el);
REDIS_NOTUSED(mask);
char buf[REDIS_IOBUF_LEN];
ssize_t nwritten, buflen;
/* Before sending the RDB file, we send the preamble as configured by the
* replication process. Currently the preamble is just the bulk count of
* the file in the form "$<length>\r\n". */
if (slave->replpreamble) {
nwritten = write(fd,slave->replpreamble,sdslen(slave->replpreamble));
if (nwritten == -1) {
redisLog(REDIS_VERBOSE,"Write error sending RDB preamble to slave: %s",
strerror(errno));
freeClient(slave);
return;
}
sdsrange(slave->replpreamble,nwritten,-1);
if (sdslen(slave->replpreamble) == 0) {
sdsfree(slave->replpreamble);
slave->replpreamble = NULL;
/* fall through sending data. */
} else {
return;
}
}
/* If the preamble was already transfered, send the RDB bulk data. */
lseek(slave->repldbfd,slave->repldboff,SEEK_SET);
// 读取 RDB 数据
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;
}
// 写入数据到 slave
if ((nwritten = write(fd,buf,buflen)) == -1) {
if (errno != EAGAIN) {
redisLog(REDIS_WARNING,"Write error sending DB to slave: %s",
strerror(errno));
freeClient(slave);
}
return;
}
// 如果写入成功,那么更新写入字节数到 repldboff ,等待下次继续写入
slave->repldboff += nwritten;
// 如果写入已经完成
if (slave->repldboff == slave->repldbsize) {
// 关闭 RDB 文件描述符
close(slave->repldbfd);
slave->repldbfd = -1;
// 删除之前绑定的写事件处理器
aeDeleteFileEvent(server.el,slave->fd,AE_WRITABLE);
// 将状态更新为 REDIS_REPL_ONLINE
slave->replstate = REDIS_REPL_ONLINE;
// 更新响应时间
slave->repl_ack_time = server.unixtime;
// 创建向从服务器发送命令的写事件处理器
// 将保存并发送 RDB 期间的回复全部发送给从服务器
if (aeCreateFileEvent(server.el, slave->fd, AE_WRITABLE,
sendReplyToClient, slave) == AE_ERR) {
redisLog(REDIS_WARNING,"Unable to register writable event for slave bulk transfer: %s", strerror(errno));
freeClient(slave);
return;
}
// 刷新低延迟 slave 数量
refreshGoodSlavesCount();
redisLog(REDIS_NOTICE,"Synchronization with slave succeeded");
}
}
6:从服务器收到RDB文件后,会为主服务器设置一个客户端用于传播同步后续的内容
// 异步 RDB 文件读取函数
#define REPL_MAX_WRITTEN_BEFORE_FSYNC (1024*1024*8) /* 8 MB */
void readSyncBulkPayload(aeEventLoop *el, int fd, void *privdata, int mask) {
char buf[4096];
ssize_t nread, readlen;
off_t left;
REDIS_NOTUSED(el);
REDIS_NOTUSED(privdata);
REDIS_NOTUSED(mask);
/* If repl_transfer_size == -1 we still have to read the bulk length
* from the master reply. */
// 读取 RDB 文件的大小
if (server.repl_transfer_size == -1) {
// 调用读函数
if (syncReadLine(fd,buf,1024,server.repl_syncio_timeout*1000) == -1) {
redisLog(REDIS_WARNING,
"I/O error reading bulk count from MASTER: %s",
strerror(errno));
goto error;
}
// 出错?
if (buf[0] == '-') {
redisLog(REDIS_WARNING,
"MASTER aborted replication with an error: %s",
buf+1);
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
* timestamp. */
// 只接到了一个作用和 PING 一样的 '\0'
// 更新最后互动时间
server.repl_transfer_lastio = server.unixtime;
return;
} else if (buf[0] != '$') {
// 读入的内容出错,和协议格式不符
redisLog(REDIS_WARNING,"Bad protocol from MASTER, the first byte is not '$' (we received '%s'), are you sure the host and port are right?", buf);
goto error;
}
// 分析 RDB 文件大小
server.repl_transfer_size = strtol(buf+1,NULL,10);
redisLog(REDIS_NOTICE,
"MASTER <-> SLAVE sync: receiving %lld bytes from master",
(long long) server.repl_transfer_size);
return;
}
/* Read bulk data */
// 读数据
// 还有多少字节要读?
left = server.repl_transfer_size - server.repl_transfer_read;
readlen = (left < (signed)sizeof(buf)) ? left : (signed)sizeof(buf);
// 读取
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;
}
// 更新最后 RDB 产生的 IO 时间
server.repl_transfer_lastio = server.unixtime;
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 synchronization: %s", strerror(errno));
goto error;
}
// 加上刚读取好的字节数
server.repl_transfer_read += nread;
/* Sync data on disk from time to time, otherwise at the end of the transfer
* we may suffer a big delay as the memory buffers are copied into the
* actual disk. */
// 定期将读入的文件 fsync 到磁盘,以免 buffer 太多,一下子写入时撑爆 IO
if (server.repl_transfer_read >=
server.repl_transfer_last_fsync_off + REPL_MAX_WRITTEN_BEFORE_FSYNC)
{
off_t sync_size = server.repl_transfer_read -
server.repl_transfer_last_fsync_off;
rdb_fsync_range(server.repl_transfer_fd,
server.repl_transfer_last_fsync_off, sync_size);
server.repl_transfer_last_fsync_off += sync_size;
}
/* Check if the transfer is now complete */
// 检查 RDB 是否已经传送完毕
if (server.repl_transfer_read == server.repl_transfer_size) {
// 完毕,将临时文件改名为 dump.rdb
if (rename(server.repl_transfer_tmpfile,server.rdb_filename) == -1) {
redisLog(REDIS_WARNING,"Failed trying to rename the temp DB into dump.rdb in MASTER <-> SLAVE synchronization: %s", strerror(errno));
replicationAbortSyncTransfer();
return;
}
// 先清空旧数据库
redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Flushing old data");
signalFlushedDb(-1);
emptyDb(replicationEmptyDbCallback);
/* 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. */
// 先删除主服务器的读事件监听,因为 rdbLoad() 函数也会监听读事件
aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
// 载入 RDB
if (rdbLoad(server.rdb_filename) != 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 */
// 关闭临时文件
zfree(server.repl_transfer_tmpfile);
close(server.repl_transfer_fd);
// 将主服务器设置成一个 redis client
// 注意 createClient 会为主服务器绑定事件,为接下来接收命令做好准备
server.master = createClient(server.repl_transfer_s);
// 标记这个客户端为主服务器
server.master->flags |= REDIS_MASTER;
// 标记它为已验证身份
server.master->authenticated = 1;
// 更新复制状态
server.repl_state = REDIS_REPL_CONNECTED;
// 设置主服务器的复制偏移量
server.master->reploff = server.repl_master_initial_offset;
// 保存主服务器的 RUN ID
memcpy(server.master->replrunid, server.repl_master_runid,
sizeof(server.repl_master_runid));
/* If master offset is set to -1, this master is old and is not
* PSYNC capable, so we flag it accordingly. */
// 如果 offset 被设置为 -1 ,那么表示主服务器的版本低于 2.8
// 无法使用 PSYNC ,所以需要设置相应的标识值
if (server.master->reploff == -1)
server.master->flags |= REDIS_PRE_PSYNC;
redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Finished with success");
/* Restart the AOF subsystem now that we finished the sync. This
* will trigger an AOF rewrite, and when done will start appending
* to the new file. */
// 如果有开启 AOF 持久化,那么重启 AOF 功能,并强制生成新数据库的 AOF 文件
if (server.aof_state != REDIS_AOF_OFF) {
int retry = 10;
// 关闭
stopAppendOnly();
// 再重启
while (retry-- && startAppendOnly() == REDIS_ERR) {
redisLog(REDIS_WARNING,"Failed enabling the AOF after successful master synchronization! Trying it again in one second.");
sleep(1);
}
if (!retry) {
redisLog(REDIS_WARNING,"FATAL: this slave instance finished the synchronization with its master, but the AOF can't be turned on. Exiting now.");
exit(1);
}
}
}
return;
error:
replicationAbortSyncTransfer();
return;
}
7:定时器操作,主服务器向从服务器发送ping,从服务器向主服务器发送ack汇报当前复制进度,避免部分传播命令丢失。
if (server.masterhost && server.master &&
!(server.master->flags & REDIS_PRE_PSYNC))
replicationSendAck();
/* If we have attached slaves, PING them from time to time.
*
* 如果服务器有从服务器,定时向它们发送 PING 。
*
* 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.
*
* 这样从服务器就可以实现显式的 master 超时判断机制,
* 即使 TCP 连接未断开也是如此。
*/
if (!(server.cronloops % (server.repl_ping_slave_period * server.hz))) {
listIter li;
listNode *ln;
robj *ping_argv[1];
/* First, send PING */
// 向所有已连接 slave (状态为 ONLINE)发送 PING
ping_argv[0] = createStringObject("PING",4);
replicationFeedSlaves(server.slaves, server.slaveseldb, ping_argv, 1);
decrRefCount(ping_argv[0]);
/* Second, send a newline to all the slaves in pre-synchronization
* stage, that is, slaves waiting for the master to create the RDB file.
*
* 向那些正在等待 RDB 文件的从服务器(状态为 BGSAVE_START 或 BGSAVE_END)
* 发送 "\n"
*
* The newline will be ignored by the slave but will refresh the
* last-io timer preventing a timeout.
*
* 这个 "\n" 会被从服务器忽略,
* 它的作用就是用来防止主服务器因为长期不发送信息而被从服务器误判为超时
*/
listRewind(server.slaves,&li);
while((ln = listNext(&li))) {
redisClient *slave = ln->value;
if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START ||
slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) {
if (write(slave->fd, "\n", 1) == -1) {
/* Don't worry, it's just a ping. */
}
}
}
}
二、哨兵机制
1.建立主服务器、从服务器、哨兵拓扑结构:
1.1:首先通过info,publish:hello建立连接,连接又分为命令连接以及订阅连接,哨兵与哨兵之间没有订阅连接,因为订阅连接的主要目的是为了让哨兵与哨兵互相得到彼此信息。
代码如下(示例):
// 如果 sentinel 与实例处于断线(未连接)状态,那么创建连向实例的异步连接。
void sentinelReconnectInstance(sentinelRedisInstance *ri) {
// 示例未断线(已连接),返回
if (!(ri->flags & SRI_DISCONNECTED)) return;
/* Commands connection. */
// 对所有实例创建一个用于发送 Redis 命令的连接
if (ri->cc == NULL) {
// 连接实例
ri->cc = redisAsyncConnect(ri->addr->ip,ri->addr->port);
// 连接出错
if (ri->cc->err) {
sentinelEvent(REDIS_DEBUG,"-cmd-link-reconnection",ri,"%@ #%s",
ri->cc->errstr);
sentinelKillLink(ri,ri->cc);
// 连接成功
} else {
// 设置连接属性
ri->cc_conn_time = mstime();
ri->cc->data = ri;
redisAeAttach(server.el,ri->cc);
// 设置连线 callback
redisAsyncSetConnectCallback(ri->cc,
sentinelLinkEstablishedCallback);
// 设置断线 callback
redisAsyncSetDisconnectCallback(ri->cc,
sentinelDisconnectCallback);
// 发送 AUTH 命令,验证身份
sentinelSendAuthIfNeeded(ri,ri->cc);
sentinelSetClientName(ri,ri->cc,"cmd");
/* Send a PING ASAP when reconnecting. */
sentinelSendPing(ri);
}
}
/* Pub / Sub */
// 对主服务器和从服务器,创建一个用于订阅频道的连接
if ((ri->flags & (SRI_MASTER|SRI_SLAVE)) && ri->pc == NULL) {
// 连接实例
ri->pc = redisAsyncConnect(ri->addr->ip,ri->addr->port);
// 连接出错
if (ri->pc->err) {
sentinelEvent(REDIS_DEBUG,"-pubsub-link-reconnection",ri,"%@ #%s",
ri->pc->errstr);
sentinelKillLink(ri,ri->pc);
// 连接成功
} else {
int retval;
// 设置连接属性
ri->pc_conn_time = mstime();
ri->pc->data = ri;
redisAeAttach(server.el,ri->pc);
// 设置连接 callback
redisAsyncSetConnectCallback(ri->pc,
sentinelLinkEstablishedCallback);
// 设置断线 callback
redisAsyncSetDisconnectCallback(ri->pc,
sentinelDisconnectCallback);
// 发送 AUTH 命令,验证身份
sentinelSendAuthIfNeeded(ri,ri->pc);
// 为客户但设置名字 "pubsub"
sentinelSetClientName(ri,ri->pc,"pubsub");
/* Now we subscribe to the Sentinels "Hello" channel. */
// 发送 SUBSCRIBE __sentinel__:hello 命令,订阅频道
retval = redisAsyncCommand(ri->pc,
sentinelReceiveHelloMessages, NULL, "SUBSCRIBE %s",
SENTINEL_HELLO_CHANNEL);
// 订阅出错,断开连接
if (retval != REDIS_OK) {
/* If we can't subscribe, the Pub/Sub connection is useless
* and we can simply disconnect it and try again. */
sentinelKillLink(ri,ri->pc);
return;
}
}
}
/* Clear the DISCONNECTED flags only if we have both the connections
* (or just the commands connection if this is a sentinel instance). */
// 如果实例是主服务器或者从服务器,那么当 cc 和 pc 两个连接都创建成功时,关闭 DISCONNECTED 标识
// 如果实例是 Sentinel ,那么当 cc 连接创建成功时,关闭 DISCONNECTED 标识
if (ri->cc && (ri->flags & SRI_SENTINEL || ri->pc))
ri->flags &= ~SRI_DISCONNECTED;
}
// 根据时间和实例类型等情况,向实例发送命令,比如 INFO 、PING 和 PUBLISH
// 虽然函数的名字包含 Ping ,但命令并不只发送 PING 命令
/* Send periodic PING, INFO, and PUBLISH to the Hello channel to
* the specified master or slave instance. */
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. */
// 函数不能在网络连接未创建时执行
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. */
// 为了避免 sentinel 在实例处于不正常状态时,发送过多命令
// sentinel 只在待发送命令的数量未超过 SENTINEL_MAX_PENDING_COMMANDS 常量时
// 才进行命令发送
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. */
// 对于从服务器来说, sentinel 默认每 SENTINEL_INFO_PERIOD 秒向它发送一次 INFO 命令
// 但是,当从服务器的主服务器处于 SDOWN 状态,或者正在执行故障转移时
// 为了更快速地捕捉从服务器的变动, sentinel 会将发送 INFO 命令的频率该为每秒一次
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;
// 实例不是 Sentinel (主服务器或者从服务器)
// 并且以下条件的其中一个成立:
// 1)SENTINEL 未收到过这个服务器的 INFO 命令回复
// 2)距离上一次该实例回复 INFO 命令已经超过 info_period 间隔
// 那么向实例发送 INFO 命令
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);
}
}
1.2:建立起拓扑连接后通过命令连接定时发送ping来获取当前主服务器是否下线,若判断主管下线,则需要进一步判断是否客观下线。
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.
*
* 如果检测到连接的活跃度(activity)很低,那么考虑重断开连接,并进行重连
*
* 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. */
// 考虑断开实例的 cc 连接
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.
*/
// 考虑断开实例的 pc 连接
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:
*
* 更新 SDOWN 标识。如果以下条件被满足,那么 Sentinel 认为实例已下线:
*
* 1) It is not replying.
* 它没有回应命令
* 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.
* Sentinel 认为实例是主服务器,这个服务器向 Sentinel 报告它将成为从服务器,
* 但在超过给定时限之后,服务器仍然没有完成这一角色转换。
*/
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) {
// 发送事件
sentinelEvent(REDIS_WARNING,"+sdown",ri,"%@");
// 记录进入 SDOWN 状态的时间
ri->s_down_since_time = mstime();
// 打开 SDOWN 标志
ri->flags |= SRI_S_DOWN;
}
} else {
// 移除(可能有的) SDOWN 状态
/* Is subjectively up */
if (ri->flags & SRI_S_DOWN) {
// 发送事件
sentinelEvent(REDIS_WARNING,"-sdown",ri,"%@");
// 移除相关标志
ri->flags &= ~(SRI_S_DOWN|SRI_SCRIPT_KILL_SENT);
}
}
}
void sentinelCheckObjectivelyDown(sentinelRedisInstance *master) {
dictIterator *di;
dictEntry *de;
int quorum = 0, odown = 0;
// 如果当前 Sentinel 将主服务器判断为主观下线
// 那么检查是否有其他 Sentinel 同意这一判断
// 当同意的数量足够时,将主服务器判断为客观下线
if (master->flags & SRI_S_DOWN) {
/* Is down for enough sentinels? */
// 统计同意的 Sentinel 数量(起始的 1 代表本 Sentinel)
quorum = 1; /* the current sentinel. */
/* Count all the other sentinels. */
// 统计其他认为 master 进入下线状态的 Sentinel 的数量
di = dictGetIterator(master->sentinels);
while((de = dictNext(di)) != NULL) {
sentinelRedisInstance *ri = dictGetVal(de);
// 该 SENTINEL 也认为 master 已下线
if (ri->flags & SRI_MASTER_DOWN) quorum++;
}
dictReleaseIterator(di);
// 如果投票得出的支持数目大于等于判断 ODOWN 所需的票数
// 那么进入 ODOWN 状态
if (quorum >= master->quorum) odown = 1;
}
/* Set the flag accordingly to the outcome. */
if (odown) {
// master 已 ODOWN
if ((master->flags & SRI_O_DOWN) == 0) {
// 发送事件
sentinelEvent(REDIS_WARNING,"+odown",master,"%@ #quorum %d/%d",
quorum, master->quorum);
// 打开 ODOWN 标志
master->flags |= SRI_O_DOWN;
// 记录进入 ODOWN 的时间
master->o_down_since_time = mstime();
}
} else {
// 未进入 ODOWN
if (master->flags & SRI_O_DOWN) {
// 如果 master 曾经进入过 ODOWN 状态,那么移除该状态
// 发送事件
sentinelEvent(REDIS_WARNING,"-odown",master,"%@");
// 移除 ODOWN 标志
master->flags &= ~SRI_O_DOWN;
}
}
}
1.3:当发生客观下线则需要进行故障转移(通过状态机机制进行状态转移),需要注意的是由于redis是单线程所以一些将从服务器提升为主服务的操作都是异步进行,避免阻塞后续定时事件的发生。
// 执行故障转移
void sentinelFailoverStateMachine(sentinelRedisInstance *ri) {
redisAssert(ri->flags & SRI_MASTER);
// master 未进入故障转移状态,直接返回
if (!(ri->flags & SRI_FAILOVER_IN_PROGRESS)) return;
switch(ri->failover_state) {
// 等待故障转移开始
case SENTINEL_FAILOVER_STATE_WAIT_START:
sentinelFailoverWaitStart(ri);
break;
// 选择新主服务器
case SENTINEL_FAILOVER_STATE_SELECT_SLAVE:
sentinelFailoverSelectSlave(ri);
break;
// 升级被选中的从服务器为新主服务器
case SENTINEL_FAILOVER_STATE_SEND_SLAVEOF_NOONE:
sentinelFailoverSendSlaveOfNoOne(ri);
break;
// 等待升级生效,如果升级超时,那么重新选择新主服务器
// 具体情况请看 sentinelRefreshInstanceInfo 函数
case SENTINEL_FAILOVER_STATE_WAIT_PROMOTION:
sentinelFailoverWaitPromotion(ri);
break;
// 向从服务器发送 SLAVEOF 命令,让它们同步新主服务器
case SENTINEL_FAILOVER_STATE_RECONF_SLAVES:
sentinelFailoverReconfNextSlave(ri);
break;
}
}
3.不解
对于redis异步连接部分尚未看懂,后续进行补充