二十一:redis主从复制实现原理

r囧r小猫 2022-12-23 06:52 118阅读 0赞

# 二十一:redis主从复制实现原理 #

## 主从复制实现原理 ##

本文使用源码redis 2.2

### 主从结构有两种结构： ###

1.一主多从
    2. 级联结果

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center]

### 复制方式分为： ###

1.全量复制
    2. 增量同步

全量复制： 发生在slave初始阶段及slave断线重新连接到master  
增量同步： 当全量复制完成后，主服务每执行一个写命令就会向从服务发送命令。

## 全量复制 ##

1.当slave启动连接到master发送sync命令  
2.master接收到sync后，开始执行bgsave命令生成RDB  
3.master后续写入命令存入到缓冲区  
4.slave与master建立连接，读取master的rdb数据  
5.slave读取数据完成，清空缓存加载rdb  
6.准备接受master的后续写命令

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center 1]

### slave端 ###

\--------------------------slave-------------------------------------------------------

#### main ####

/** * file: redis.c * function: main() * 这里展示的仅是main的部分，为了排除干扰仅说明主从复制的原理 **/
    int main(int argc, char **argv) { 
    // 初始化server配置（未读取配置文件）
    initServerConfig();
    //启动参数 redis-server /path/redis.conf
    if (argc == 2) { 
            //加载配置文件
            loadServerConfig(argv[1]);
        }
    
    }

#### loadServerConfig ####

/** * file: config.c * function: loadServerConfig() **/
    void loadServerConfig(char *filename){ 
        // 如果当前启动的是slave，则配置文件有配置master的ip、port （slaveof是redis5.0之前的的配置，5.0后使用replicaof ）
        if (!strcasecmp(argv[0],"slaveof") && argc == 3) { 
            server.masterhost = sdsnew(argv[1]);
            server.masterport = atoi(argv[2]);
            //设置复制状态为可连接
            server.replstate = REDIS_REPL_CONNECT;
        }
    }

#### replicationCron ####

/** * file: replication.c * function: replicationCron() * des: serverCronding定时任务，调用replicationCron定时复制任务，通常当重连接到master或传输失败时执行 **/
    void replicationCron(void) { 
        // 上面的初始化读取配置文件设置了replstate为REDIS_REPL_CONNECT
        if (server.replstate == REDIS_REPL_CONNECT) { 
            redisLog(REDIS_NOTICE,"Connecting to MASTER...");
            // 发送sync到master，master准备bgsave生成rdb文件
            if (syncWithMaster() == REDIS_OK) { 
                redisLog(REDIS_NOTICE,"MASTER <-> SLAVE sync started: SYNC sent");
            }
        }
    }

#### syncWithMaster ####

/** * replication.c * * syncWithMaster() **/
    int syncWithMaster(void) { 
        char buf[1024], tmpfile[256], authcmd[1024];
        // tcp socket连接master
        int fd = anetTcpConnect(NULL,server.masterhost,server.masterport);
        // dfd： maxtries: 打开rdb文件最大的重试次数
        int dfd, maxtries = 5;
        // 不能连接到master
        if (fd == -1) { 
            redisLog(REDIS_WARNING,"Unable to connect to MASTER: %s",
                strerror(errno));
            return REDIS_ERR;
        }
    
        //如果有配置master的密码则尝试认证
        if(server.masterauth) { 
            // 构造 auth password 命令
        	snprintf(authcmd, 1024, "AUTH %s\r\n", server.masterauth);
            //阻塞式的执行认证命令，并且设置超时时间5秒
        	if (syncWrite(fd, authcmd, strlen(server.masterauth)+7, 5) == -1) { 
                close(fd);
                redisLog(REDIS_WARNING,"Unable to AUTH to MASTER: %s",
                    strerror(errno));
                return REDIS_ERR;
        	}
           // 阻塞式的读取auth password执行的结果，并设置超时间时间3600秒
            if (syncReadLine(fd,buf,1024,3600) == -1) { 
                close(fd);
                redisLog(REDIS_WARNING,"I/O error reading auth result from MASTER: %s",
                    strerror(errno));
                return REDIS_ERR;
            }
            //resp协议,五种数据类型，这里执行auth命令可能的结果前缀+、-; 如果不是+开头，则表示密码错误
            if (buf[0] != '+') { 
                close(fd);
                redisLog(REDIS_WARNING,"Cannot AUTH to MASTER, is the masterauth password correct?");
                return REDIS_ERR;
            }
        }
    
        //开始准备执行sync命令，结果返回-1表示执行失败
        if (syncWrite(fd,"SYNC \r\n",7,5) == -1) { 
            close(fd);
            redisLog(REDIS_WARNING,"I/O error writing to MASTER: %s",
                strerror(errno));
            return REDIS_ERR;
        }
    
        //最大尝试5次，打开文件
        while(maxtries--) { 
            // 构造临时文件名
            snprintf(tmpfile,256,
                "temp-%d.%ld.rdb",(int)time(NULL),(long int)getpid());
            //O_CREAT： 如果指定文件不存在，则创建这个文件
            //O_WRONLY：只写模式
            //O_EXCL： 如果要创建的文件已存在，则返回-1，并且修改error的值
            //oepn函数返回值成功则返回文件描述符，否则返回-1
            dfd = open(tmpfile,O_CREAT|O_WRONLY|O_EXCL,0644);
            // 创建成功，直接推出循环
            if (dfd != -1) break;
            //否则睡眠1秒，继续尝试创建文件，最多尝试5次。
            sleep(1);
        }
        //如果创建文件失败，输出错误信息，结束函数
        if (dfd == -1) { 
            close(fd);
            redisLog(REDIS_WARNING,"Opening the temp file needed for MASTER <-> SLAVE synchronization: %s",strerror(errno));
            return REDIS_ERR;
        }
    
        // 从master非阻塞的下载rdb文件, readSyncBulkPayload读master的rdb数据
        if (aeCreateFileEvent(server.el, fd, AE_READABLE, readSyncBulkPayload, NULL)
                == AE_ERR)
        { 
            close(fd);
            redisLog(REDIS_WARNING,"Can't create readable event for SYNC");
            return REDIS_ERR;
        }
        server.replstate = REDIS_REPL_TRANSFER;
        server.repl_transfer_left = -1;
        server.repl_transfer_s = fd;
        server.repl_transfer_fd = dfd;
        server.repl_transfer_lastio = time(NULL);
        server.repl_transfer_tmpfile = zstrdup(tmpfile);
        return REDIS_OK;
    }

#### readSyncBulkPayload ####

/** * 实际读取master rdb到slave的函数 ** file: replication.c ** function: readSyncBulkPayload **/
    /* 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);
    
       // 从master读取的字节大小,如果等于-1，则读取
        if (server.repl_transfer_left == -1) { 
            //读取master rdb一行字符
            if (syncReadLine(fd,buf,1024,3600) == -1) { 
                redisLog(REDIS_WARNING,
                    "I/O error reading bulk count from MASTER: %s",
                    strerror(errno));
                replicationAbortSyncTransfer();
                return;
            }
            // 失败的命令，直接返回,关闭与master的连接
            if (buf[0] == '-') { 
                redisLog(REDIS_WARNING,
                    "MASTER aborted replication with an error: %s",
                    buf+1);
                replicationAbortSyncTransfer();
                return;
            // 读到行末，继续读取下一行
            } else if (buf[0] == '\0') { 
                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;
            }
            // 计算需要读取的字节数
            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;
        }
    
        // 计算准备读取的字节长度,如果repl_transfer_left小于buf，则读repl_transfer_left大小
        readlen = (server.repl_transfer_left < (signed)sizeof(buf)) ?
            server.repl_transfer_left : (signed)sizeof(buf);
        //读取数据从master
        nread = read(fd,buf,readlen);
        //读失败，关闭连接
        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);
        //写master读到的数据到slave的repl_transfer_fd（temp文件）
        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;
        // 检查是否已经读完了master的rdb数据
        if (server.repl_transfer_left == 0) { 
            //重命名slave tmp文件名为slave rdb文件名
            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;
            }
            //输出日志,在master与salve之间执行sync命令完成，
            redisLog(REDIS_NOTICE, "MASTER <-> SLAVE sync: Loading DB in memory");
            //清空slave的缓存数据
            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. */
             //关闭事件连接s
            aeDeleteFileEvent(server.el,server.repl_transfer_s,AE_READABLE);
            //加载rdb到redis
            if (rdbLoad(server.dbfilename) != REDIS_OK) { 
                redisLog(REDIS_WARNING,"Failed trying to load the MASTER synchronization DB from disk");
                replicationAbortSyncTransfer();
                return;
            }
           //释放临时文件、关闭临时文件的连接、slave创建连接，准备接受master后续的写命令
            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");
            /* Rewrite the AOF file now that the dataset changed. */
            if (server.appendonly) rewriteAppendOnlyFileBackground();
        }
    }

### master ###

\---------------------master-----------------------------------------------------------------

#### syncCommand ####

/** * master * file: replication.c * funtion: syncCommand **/
    void syncCommand(redisClient *c) { 
        //slave不执行sync命令
        if (c->flags & REDIS_SLAVE) return;
    
        // 这两个参数配置，说明是自己slave，不执行sync命令
        if (server.masterhost && server.replstate != REDIS_REPL_CONNECTED) { 
            addReplyError(c,"Can't SYNC while not connected with my master");
            return;
        }
         //当slave发送sync表示slave启动重新连接到master，重新全量复制mater的数据
         //所以执行sync之前，数据缓存区应该是没有数据的。
        if (listLength(c->reply) != 0) { 
            addReplyError(c,"SYNC is invalid with pending input");
            return;
        }
        //master输出日志，准备全量复制数据到slave
        redisLog(REDIS_NOTICE,"Slave ask for synchronization");
       // bgsave的子进程 != -1表示子进程正在保存快照中
        if (server.bgsavechildpid != -1) { 
           // 如果bgsave正在运行，则判断是否有其他slave执行了sync，如果有那么就等待
           //并当前连接和那个slave保持一致.
            redisClient *slave;
            listNode *ln;
            listIter li;
            //获取当前master连接的所有slave的迭代器
            listRewind(server.slaves,&li);
            while((ln = listNext(&li))) { 
                //迭代，如果有其他的slave在等待bgsave结束，则返回这个slave
                //否则返回空的listnode
                slave = ln->value;
                if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_END) break;
            }
            if (ln) { 
              //表示有其他的slava执行了bgsave
              //当前连接salve，复制那个slave缓冲区、以及复制状态设为等待bgsave
                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 { 
                //没有其他salve执行bgsave表示，配置的save参数(表示在多少秒内直到有多少key改变才触发rdb持久化)
                //生效了，设置salve的复制状态为bgsave
                c->replstate = REDIS_REPL_WAIT_BGSAVE_START;
                redisLog(REDIS_NOTICE,"Waiting for next BGSAVE for SYNC");
            }
        } else { 
           //没有其他salve执行bgsave,save的配置没有触发bgsave的条件，
           //那么当前连接slave则触发bgsave
            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;
            }
            //设置salve连接的复制状态为`等待bgsave结束`
            c->replstate = REDIS_REPL_WAIT_BGSAVE_END;
        }
        c->repldbfd = -1;
        c->flags |= REDIS_SLAVE;
        c->slaveseldb = 0;
        //当前salve连接加入到master的slave队列中
        listAddNodeTail(server.slaves,c);
        return;
    }

#### updateSlavesWaitingBgsave ####

/** * file: replication.c * function: updateSlavesWaitingBgsave * * des: redis.c的serverCron函数会定时执行,当bgsave执行完毕 * 会开始执行backgroundSaveDoneHandler函数，调用updateSlavesWaitingBgsave * 发送rdb文件的数据到salve * **/
    //当bgsave成功，传入的bgsaveerr= REDIS_OK
    void updateSlavesWaitingBgsave(int bgsaveerr) { 
        listNode *ln;
        int startbgsave = 0;
        listIter li;
        // 获取li,为连接master的slave的迭代器
        listRewind(server.slaves,&li);
        while((ln = listNext(&li))) { 
            //当前迭代slave的连接对象
            redisClient *slave = ln->value;
            //如果slave的复制状态是等待bgsave则设置为bgsave结束了
            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) { 
                //如果slave的复制状态是 等待bgsave结束，则开始发送master的rdb数据
                //到slave
                struct redis_stat buf;
                //如果bgsave是失败的则，释放slave的连接。（slave后面后再次发起sync请求）
                if (bgsaveerr != REDIS_OK) { 
                    freeClient(slave);
                    redisLog(REDIS_WARNING,"SYNC failed. BGSAVE child returned an error");
                    continue;
                }
                //否则，bgsave是成功的，以只读方式打开master的rdb
                if ((slave->repldbfd = open(server.dbfilename,O_RDONLY)) == -1 ||
                    redis_fstat(slave->repldbfd,&buf) == -1) { 
                    //打开rdb文件失败，则释放slave的连接
                    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);
                //使用redis自己的事件驱动来写数据到slave
                //sendBulkToSlave函数为写数据的主要逻辑
                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);
                }
            }
        }
    }

## 增量复制 ##

增量复制：当全量复制完成后，主服务每执行一个写命令就会向从服务发送命令。

以set时判断key是否过期为例，分析增量复制：setGenericCommand —>expireIfNeeded

![在这里插入图片描述][watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center 2]

### master ###

#### propagateExpire ####

/** * file: db.c * function： propagateExpire * * dec: 在expireIfNeeded函数中，当判断key是否过期，如果key过期了则调用 * propagateExpire函数传播删除key，到slave及aof文件中 **/
    void propagateExpire(redisDb *db, robj *key) { 
        robj *argv[2];
    
        argv[0] = createStringObject("DEL",3);
        argv[1] = key;
        incrRefCount(key);
        //如果开启了aof，则写入命令到aof文件中
        if (server.appendonly)
            feedAppendOnlyFile(server.delCommand,db->id,argv,2);
        if (listLength(server.slaves))
            //如果master有连接的slave，则传播del命令到所有的slave
            replicationFeedSlaves(server.slaves,db->id,argv,2);
    
        decrRefCount(argv[0]);
        decrRefCount(argv[1]);
    }

#### replicationFeedSlaves ####

/** * file: replication.c * function: replicationFeedSlaves * * des: 复制命令到slave **/
    void replicationFeedSlaves(list *slaves, int dictid, robj **argv, int argc) { 
        listNode *ln;
        listIter li;
        int outc = 0, j;
        robj **outv;
        //构造存放发送命令的数组
        // REDIS_STATIC_ARGS = 8 
        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));
        }
    
        //构造消息头(根据resp协议)，* :表示后面是数组，以及指定的数组的大小
        lenobj = createObject(REDIS_STRING,
                sdscatprintf(sdsempty(), "*%d\r\n", argc));
        lenobj->refcount = 0;
        outv[outc++] = lenobj;
        //argv存放到out数组中
        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;
        }
    
    
        for (j = 0; j < outc; j++) incrRefCount(outv[j]);
        // slave的迭代器
        listRewind(slaves,&li);
        while((ln = listNext(&li))) { 
            redisClient *slave = ln->value;
    
            //如果slave的复制状态是等待开始，表示刚连接上，等待bgsave
            //全量复制还未完成，则不进行增量复制
            if (slave->replstate == REDIS_REPL_WAIT_BGSAVE_START) continue;
    
           // 选择master的数据库，如果slave与master的数据库不一样
           //则构造命令执行指定slave的数据库
            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;
            }
            // 发送数组的数据到slave的socket到slave
            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);
    }

## 参考 ##

https://www.cnblogs.com/fnlingnzb-learner/p/7040726.html

https://www.cnblogs.com/daofaziran/p/10978628.html

[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center]: /images/20221120/92eb769d5c054ccc876dd7256387b3a9.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center 1]: /images/20221120/ed8f56633a874f03a9bceb25cb11b04b.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM_size_16_color_FFFFFF_t_70_pic_center 2]: https://img-blog.csdnimg.cn/20201123144047295.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3UwMTM1NjUxNjM=,size_16,color_FFFFFF,t_70#pic_center