作者：京东物流 宫丙来
一、主从复制概述
二、CommitLog消息复制
2.1、整体概述
CommitLog主从复制的流程如下：
1.Producer发送消息到Broker Master，Broker进行消息存储，并调用handleHA进行主从同步； 2.如果是同步复制的话，参考2.6章节的同步复制； 3.如果是异步复制的话，流程如下：
1. Broker Master启动，并在指定端口监听；
2. Broker Slave启动，主动连接Broker Master，通过JAVA NIO建立TCP连接；
3. Broker Slave以每隔5s的间隔时间向服务端拉取消息，如果是第一次拉取的话，先获取本地CommitLog文件中最大的偏移量，以该偏移量向服务端拉取消息
4. Broker Master 解析请求，并返回数据给Broker Slave；
5.Broker Slave收到一批消息后，将消息写入本地CommitLog文件中，然后向Master汇报拉取进度，并更新下一次待拉取偏移量；
我们先看下异步复制的整体流程，最后再看下同步复制的流程，异步复制的入口为HAService.start();
public void start() throws Exception {
//broker master启动，接收slave请求，并处理
this.acceptSocketService.beginAccept();
this.acceptSocketService.start();
//同步复制线程启动
this.groupTransferService.start();
//broker slave启动
this.haClient.start();
}
下面分别对上面的每一步做详细说明。
2.2、HAService Master启动
public void beginAccept() throws Exception {
this.serverSocketChannel = ServerSocketChannel.open();
this.selector = RemotingUtil.openSelector();
this.serverSocketChannel.socket().setReuseAddress(true);
this.serverSocketChannel.socket().bind(this.socketAddressListen);
this.serverSocketChannel.configureBlocking(false);
this.serverSocketChannel.register(this.selector, SelectionKey.OP_ACCEPT);
}
在beginAccept方法中主要创建了ServerSocketChannel、Selector、设置TCP reuseAddress、绑定监听端口、设置为非阻塞模式，并注册OP_ACCEPT(连接事件)。可以看到在这里是通过Java原生的NIO来实现的，并没有通.NETty框架来实现。
acceptSocketService.start()启动方法代码如下：
while (!this.isStopped()) {
try {
//获取事件
this.selector.select(1000);
Set<SelectionKey> selected = this.selector.selectedKeys();
if (selected != null) {
for (SelectionKey k : selected) {
//处理OP_ACCEPT事件，并创建HAConnection
if ((k.readyOps() & SelectionKey.OP_ACCEPT) != 0) {
SocketChannel sc = ((ServerSocketChannel) k.channel()).accept();
if (sc != null) {
HAConnection conn = new HAConnection(HAService.this, sc);
//主要是启动readSocketService，writeSocketService这两个线程
conn.start();
HAService.this.addConnection(conn);
}
}
}
selected.clear();
}
} catch (Exception e) {
log.error(this.getServiceName() + " service has exception.", e);
}
}
选择器每1s处理一次处理一次连接就绪事件。连接事件就绪后，调用ServerSocketChannel的accept()方法创建SocketChannel，与服务端数据传输的通道。然后为每一个连接创建一个HAConnection对象，该HAConnection将负责Master-Slave数据同步逻辑。HAConnection.start方法如下：
public void start() {
this.readSocketService.start();
this.writeSocketService.start();
}
2.3、HAClient启动
while (!this.isStopped()) {
try {
//和broker master建立连接，通过java nio来实现
if (this.connectMaster()) {
//在心跳的同时，上报offset
if (this.isTimeToReportOffset()) {
//上报offset
boolean result = this.reportSlaveMaxOffset(this.currentReportedOffset);
if (!result) {
this.closeMaster();
}
}
this.selector.select(1000);
//处理网络读请求，也就是处理从Master传回的消息数据
boolean ok = this.processReadEvent();
if (!ok) {
this.closeMaster();
}
if (!reportSlaveMaxOffsetPlus()) {
continue;
}
long interval =
HAService.this.getDefaultMessageStore().getSystemClock().now()
- this.lastWriteTimestamp;
if (interval > HAService.this.getDefaultMessageStore().getMessageStoreConfig()
.getHaHousekeepingInterval()) {
log.warn("HAClient, housekeeping, found this connection[" + this.masterAddress
+ "] expired, " + interval);
this.closeMaster();
log.warn("HAClient, master not response some time, so close connection");
}
} else {
this.wAItForRunning(1000 * 5);
}
} catch (Exception e) {
log.warn(this.getServiceName() + " service has exception. ", e);
this.waitForRunning(1000 * 5);
}
}
2.3.1、HAService主从建立连接
如果socketChannel为空，则尝试连接Master，如果Master地址为空，返回false。
private boolean connectMaster() throws ClosedChannelException {
if (null == socketChannel) {
String addr = this.masterAddress.get();
if (addr != null) {
SocketAddress socketAddress = RemotingUtil.string2SocketAddress(addr);
if (socketAddress != null) {
this.socketChannel = RemotingUtil.connect(socketAddress);
if (this.socketChannel != null) {
//注册读事件，监听broker master返回的数据
this.socketChannel.register(this.selector, SelectionKey.OP_READ);
}
}
}
//获取当前的offset
this.currentReportedOffset = HAService.this.defaultMessageStore.getMaxPhyOffset();
this.lastWriteTimestamp = System.currentTimeMillis();
}
return this.socketChannel != null;
}
Broker Slave通过NIO来进行Broker Master连接，代码如下：
SocketChannel sc = null;
sc = SocketChannel.open();
sc.configureBlocking(true);
sc.socket().setSoLinger(false, -1);
sc.socket().setTcpNoDelay(true);
sc.socket().setReceiveBufferSize(1024 * 64);
sc.socket().setSendBufferSize(1024 * 64);
sc.socket().connect(remote, timeoutMillis);
sc.configureBlocking(false);
public long getMaxPhyOffset() {
return this.commitLog.getMaxOffset();
}
public long getMaxOffset() {
return this.mAppedFileQueue.getMaxOffset();
}
public long getMaxOffset() {
MappedFile mappedFile = getLastMappedFile();
if (mappedFile != null) {
return mappedFile.getFileFromOffset() + mappedFile.getReadPosition();
}
return 0;
}
可以看到最终还是通过读取MappedFile的position来获取从的offset。
2.3.2、上报offset时间判断
private boolean isTimeToReportOffset() {
//当前时间-上次写的时间
long interval =
HAService.this.defaultMessageStore.getSystemClock().now() - this.lastWriteTimestamp;
boolean needHeart = interval > HAService.this.defaultMessageStore.getMessageStoreConfig()
.getHaSendHeartbeatInterval();
return needHeart;
}
判断逻辑为当前时间-上次写的时间>haSendHeartbeatInterval时，则进行心跳和offset的上报。haSendHeartbeatInterval默认为5s，可配置。
2.3.3、上报offset
private boolean reportSlaveMaxOffset(final long maxOffset) {
this.reportOffset.position(0);
this.reportOffset.limit(8);
this.reportOffset.putLong(maxOffset);
this.reportOffset.position(0);
this.reportOffset.limit(8);
//最多发送三次，reportOffset是否有剩余
for (int i = 0; i < 3 && this.reportOffset.hasRemaining(); i++) {
try {
this.socketChannel.write(this.reportOffset);
} catch (IOException e) {
log.error(this.getServiceName()
+ "reportSlaveMaxOffset this.socketChannel.write exception", e);
return false;
}
}
return !this.reportOffset.hasRemaining();
}
主要还是通过NIO发送请求。
2.4、Broker Master处理请求
在主从建立连接时创建了HAConnection对象，该对象主要包含了如下两个重要的线程服务类：
//负责写，将commitlog数据发送到从
private WriteSocketService writeSocketService;
//负责读，读取从上报的offset，并根据offset从Broker Master读取commitlog
private ReadSocketService readSocketService;
2.4.1、ReadSocketService接收读请求
readSocketService.run方法如下：
while (!this.isStopped()) {
try {
this.selector.select(1000);
//处理读事件
boolean ok = this.processReadEvent();
if (!ok) {
HAConnection.log.error("processReadEvent error");
break;
}
long interval = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now() - this.lastReadTimestamp;
if (interval > HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaHousekeepingInterval()) {
log.warn("ha housekeeping, found this connection[" + HAConnection.this.clientAddr + "] expired, " + interval);
break;
}
} catch (Exception e) {
HAConnection.log.error(this.getServiceName() + " service has exception.", e);
break;
}
}
processReadEvent的逻辑如下：
int readSize = this.socketChannel.read(this.byteBufferRead);
if (readSize > 0) {
readSizeZeroTimes = 0;
this.lastReadTimestamp = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now();
if ((this.byteBufferRead.position() - this.processPostion) >= 8) {
int pos = this.byteBufferRead.position() - (this.byteBufferRead.position() % 8);
//获取slave 请求的offset
long readOffset = this.byteBufferRead.getLong(pos - 8);
this.processPostion = pos;
HAConnection.this.slaveAckOffset = readOffset;
if (HAConnection.this.slaveRequestOffset < 0) {
HAConnection.this.slaveRequestOffset = readOffset;
log.info("slave[" + HAConnection.this.clientAddr + "] request offset " + readOffset);
}
//如果是同步复制的话，判断请求的offset是否push2SlaveMaxOffset相同，相同的话则唤醒master GroupTransferService
HAConnection.this.haService.notifyTransferSome(HAConnection.this.slaveAckOffset);
}
}
可以看到processReadEvent逻辑很简单，就是从ByteBuffer中解析出offset，然后设置
HAConnection.this.slaveRequestOffset；
2.4.2、WriteSocketService进行写处理
Broker Master通过
HAConnection.WriteSocketService进行CommitLog的读取，run方法主逻辑如下：
this.selector.select(1000);
//nextTransferFromWhere下次传输commitLog的起始位置
if (-1 == this.nextTransferFromWhere) {
if (0 == HAConnection.this.slaveRequestOffset) {
long masterOffset = HAConnection.this.haService.getDefaultMessageStore().getCommitLog().getMaxOffset();
masterOffset =
masterOffset
- (masterOffset % HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig()
.getMapedFileSizeCommitLog());
if (masterOffset < 0) {
masterOffset = 0;
}
this.nextTransferFromWhere = masterOffset;
} else {
this.nextTransferFromWhere = HAConnection.this.slaveRequestOffset;
}
log.info("master transfer data from " + this.nextTransferFromWhere + " to slave[" + HAConnection.this.clientAddr
+ "], and slave request " + HAConnection.this.slaveRequestOffset);
}
//获取commitLog数据
SelectMappedBufferResult selectResult = HAConnection.this.haService.getDefaultMessageStore().getCommitLogData(this.nextTransferFromWhere);
//获取commitLog数据
SelectMappedBufferResult selectResult =
HAConnection.this.haService.getDefaultMessageStore().getCommitLogData(this.nextTransferFromWhere);
if (selectResult != null) {
int size = selectResult.getSize();
if (size > HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaTransferBatchSize()) {
size = HAConnection.this.haService.getDefaultMessageStore().getMessageStoreConfig().getHaTransferBatchSize();
}
long thisOffset = this.nextTransferFromWhere;
this.nextTransferFromWhere += size;
selectResult.getByteBuffer().limit(size);
this.selectMappedBufferResult = selectResult;
// Build Header
this.byteBufferHeader.position(0);
this.byteBufferHeader.limit(headerSize);
this.byteBufferHeader.putLong(thisOffset);
this.byteBufferHeader.putInt(size);
this.byteBufferHeader.flip();
//nio发送commitlog
this.lastWriteOver = this.transferData();
} else {
//如果没有获取到commitLog数据，等待100ms
HAConnection.this.haService.getWaitNotifyObject().allWaitForRunning(1
这里面主要包括获取CommitLog数据、发送CommitLog数据这两个步骤。
2.4.2.1、获取CommitLog数据
public SelectMappedBufferResult getData(final long offset, final boolean returnFirstOnNotFound) {
int mappedFileSize = this.defaultMessageStore.getMessageStoreConfig().getMapedFileSizeCommitLog();
MappedFile mappedFile = this.mappedFileQueue.findMappedFileByOffset(offset, returnFirstOnNotFound);
if (mappedFile != null) {
int pos = (int) (offset % mappedFileSize);
SelectMappedBufferResult result = mappedFile.selectMappedBuffer(pos);
return result;
}
return null;
}
public SelectMappedBufferResult selectMappedBuffer(int pos) {
int readPosition = getReadPosition();
if (pos < readPosition && pos >= 0) {
if (this.hold()) {
ByteBuffer byteBuffer = this.mappedByteBuffer.slice();
byteBuffer.position(pos);
int size = readPosition - pos;
ByteBuffer byteBufferNew = byteBuffer.slice();
byteBufferNew.limit(size);
return new SelectMappedBufferResult(this.fileFromOffset + pos, byteBufferNew, size, this);
}
}
return null;
}
可以看到最终还是根据offset从MappedFile读取数据。
2.4.2.2、发送CommitLog数据
数据主要包括header、body两部分，数据发送的话还是通过NIO来实现，主要代码如下：
// Build Header
this.byteBufferHeader.position(0);
this.byteBufferHeader.limit(headerSize);
this.byteBufferHeader.putLong(thisOffset);
this.byteBufferHeader.putInt(size);
this.byteBufferHeader.flip();
int writeSize = this.socketChannel.write(this.byteBufferHeader);
// Write Body
if (!this.byteBufferHeader.hasRemaining()) {
while (this.selectMappedBufferResult.getByteBuffer().hasRemaining()) {
int writeSize = this.socketChannel.write(this.selectMappedBufferResult.getByteBuffer());
if (writeSize > 0) {
writeSizeZeroTimes = 0;
this.lastWriteTimestamp = HAConnection.this.haService.getDefaultMessageStore().getSystemClock().now();
} else if (writeSize == 0) {
if (++writeSizeZeroTimes >= 3) {
break;
}
} else {
throw new Exception("ha master write body error < 0");
}
}
}
CommitLog主从发送完成后，Broker Slave则会监听读事件、获取CommitLog数据，并进行CommitLog的写入。
2.5、HAClient processReadEvent
在主从建立连接后，从注册了可读事件，目的就是读取从Broker Master返回的CommitLog数据，对应的方法为HAClient.processReadEvent：
int readSize = this.socketChannel.read(this.byteBufferRead);
if (readSize > 0) {
lastWriteTimestamp = HAService.this.defaultMessageStore.getSystemClock().now();
readSizeZeroTimes = 0;
boolean result = this.dispatchReadRequest();
if (!result) {
log.error("HAClient, dispatchReadRequest error");
return false;
}
}
dispatchReadRequest方法如下：
//读取返回的body data
byte[] bodyData = new byte[bodySize];
this.byteBufferRead.position(this.dispatchPostion + msgHeaderSize);
this.byteBufferRead.get(bodyData);
HAService.this.defaultMessageStore.appendToCommitLog(masterPhyOffset, bodyData);
this.byteBufferRead.position(readSocketPos);
this.dispatchPostion += msgHeaderSize + bodySize;
//上报从的offset
if (!reportSlaveMaxOffsetPlus()) {
return false;
里面的核心逻辑主要包括如下三个步骤：
public boolean appendToCommitLog(long startOffset, byte[] data) {
//将数据写到commitlog，同普通消息的存储
boolean result = this.commitLog.appendData(startOffset, data);
//唤醒reputMessageService，构建consumeQueue,index
this.reputMessageService.wakeup();
return result;
}
2.6、同步复制
上面主要介绍了Broker的异步复制，下面再来看下Broker的同步复制的实现。同步复制的整体流程图如下：
大概说明如下：
对应的代码入口为CommitLog.handleHA方法。
public void handleHA(AppendMessageResult result, PutMessageResult putMessageResult, MessageExt messageExt) {
//如果是broker主，并且是同步复制的话
if (BrokerRole.SYNC_MASTER == this.defaultMessageStore.getMessageStoreConfig().getBrokerRole()) {
//获取HAService
HAService service = this.defaultMessageStore.getHaService();
//获取Message上的MessageConst.PROPERTY_WAIT_STORE_MSG_OK，默认是需要等待主从复制完成
if (messageExt.isWaitStoreMsgOK()) {
/**
* 判断从是否可用，判断的逻辑是：（主offset-push2SlaveMaxOffset<1024 * 1024 * 256），也就是如果主从的offset差的太多，
* 则认为从不可用， Tell the producer, slave not available
* 这里的result = mappedFile.appendMessage(msg, this.appendMessageCallback);
*/
if (service.isSlaveOK(result.getWroteOffset() + result.getWroteBytes())) {
//组装GroupCommitRequest，nextOffset=result.getWroteOffset() + result.getWroteBytes()，这里的nextOffset指的就是从要写到的offset
GroupCommitRequest request = new GroupCommitRequest(result.getWroteOffset() + result.getWroteBytes());
/**
* 调用的是this.groupTransferService.putRequest(request);将request放到requestsWrite list中。
* HAService持有GroupTransferService groupTransferService引用;
*/
service.putRequest(request);
/**
* 唤醒的是WriteSocketService，查询commitLog数据，然后发送到从。
* 在WriteSocketService获取commitLog时，如果没有获取到commitLog数据，等待100ms
* HAConnection.this.haService.getWaitNotifyObject().allWaitForRunning(100);
* 所以当commitLog新写入数据的时候，会唤醒WriteSocketService，然后查询commitLog数据，发送到从。
*/
service.getWaitNotifyObject().wakeupAll();
//等待同步复制完成，判断逻辑是： HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
boolean flushOK =
request.waitForFlush(this.defaultMessageStore.getMessageStoreConfig().getSyncFlushTimeout());
//如果同步复制失败的话，设置putMessageResult中的状态为同步从超时
if (!flushOK) {
log.error("do sync transfer other node, wait return, but failed, topic: " + messageExt.getTopic() + " tags: "
+ messageExt.getTags() + " client address: " + messageExt.getBornHostNameString());
putMessageResult.setPutMessageStatus(PutMessageStatus.FLUSH_SLAVE_TIMEOUT);
}
}
// Slave problem
else {
// Tell the producer, slave not available
putMessageResult.setPutMessageStatus(PutMessageStatus.SLAVE_NOT_AVAILABLE);
}
}
}
2.6.1、GroupTransferService启动
在HAService启动的时候，启动了GroupTransferService线程，代码如下：
public void run() {
while (!this.isStopped()) {
this.waitForRunning(10);
this.doWaitTransfer();
}
}
private void doWaitTransfer() {
synchronized (this.requestsRead) {
if (!this.requestsRead.isEmpty()) {
for (CommitLog.GroupCommitRequest req : this.requestsRead) {
/**
* req.getNextOffset：result.getWroteOffset() + result.getWroteBytes()
* push2SlaveMaxOffset：
*/
boolean transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
//在这循环5次，最多等待5s，因为slave 心跳间隔默认5s
for (int i = 0; !transferOK && i < 5; i++) {
this.notifyTransferObject.waitForRunning(1000);
transferOK = HAService.this.push2SlaveMaxOffset.get() >= req.getNextOffset();
}
if (!transferOK) {
log.warn("transfer messsage to slave timeout, " + req.getNextOffset());
}
//主从复制完成，唤醒handleHA后续操作
req.wakeupCustomer(transferOK);
}
this.requestsRead.clear();
}
}
}
wakeupCustomer：
public void wakeupCustomer(final boolean flushOK) {
this.flushOK = flushOK;
this.countDownLatch.countDown();
}
2.6.2、唤醒WriteSocketService
service.getWaitNotifyObject().wakeupAll();
唤醒的是WriteSocketService，查询commitLog数据，然后发送到从。在WriteSocketService获取commitLog时，如果没有获取到commitLog数据，等待100ms。
HAConnection.this.haService.getWaitNotifyObject().allWaitForRunning(100);所以当commitLog新写入数据的时候，会唤醒WriteSocketService，然后查询commitLog数据，发送到从。
2.6.3、同步等待，直到复制完成
boolean flushOK =
request.waitForFlush(this.defaultMessageStore.getMessageStoreConfig().getSyncFlushTimeout());
public boolean waitForFlush(long timeout) {
try {
//等待同步复制完成
this.countDownLatch.await(timeout, TimeUnit.MILLISECONDS);
return this.flushOK;
} catch (InterruptedException e) {
log.error("Interrupted", e);
return false;
}
}
}
三、元数据的复制
broker元数据的复制，主要包括topicConfig、consumerOffset、delayOffset、subscriptionGroup这几部分，整体流程图如下：
从broker通过单独的线程，每隔10s进行一次元数据的复制 ，代码入口为：BrokerController.start -> SlaveSynchronize.syncAll：
slaveSyncFuture = this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {
@Override
public void run() {
try {
//10s 进行一次主从同步
BrokerController.this.slaveSynchronize.syncAll();
}
catch (Throwable e) {
log.error("ScheduledTask SlaveSynchronize syncAll error.", e);
}
}
}, 1000 * 3, 1000 * 10, TimeUnit.MILLISECONDS);
public void syncAll() {
this.syncTopicConfig();
this.syncConsumerOffset();
this.syncDelayOffset();
this.syncSubscriptionGroupConfig();
}
3.1、syncTopicConfig
//从Master获取TopicConfig信息，最终调用的是AdminBrokerProcessor.getAllTopicConfig
TopicConfigSerializeWrapper topicWrapper =
this.brokerController.getBrokerOuterAPI().getAllTopicConfig(masterAddrBak);
if (!this.brokerController.getTopicConfigManager().getDataVersion()
.equals(topicWrapper.getDataVersion())) {
this.brokerController.getTopicConfigManager().getDataVersion()
.assignNewOne(topicWrapper.getDataVersion());
this.brokerController.getTopicConfigManager().getTopicConfigTable().clear();
this.brokerController.getTopicConfigManager().getTopicConfigTable()
.putAll(topicWrapper.getTopicConfigTable());
//将topicConfig进行持久化，对应的文件为topics.json
this.brokerController.getTopicConfigManager().persist();
log.info("Update slave topic config from master, {}", masterAddrBak)
3.2、syncConsumerOffset
//从"主Broker"获取ConsumerOffset
ConsumerOffsetSerializeWrapper offsetWrapper =
this.brokerController.getBrokerOuterAPI().getAllConsumerOffset(masterAddrBak);
//设置从的offsetTable
this.brokerController.getConsumerOffsetManager().getOffsetTable()
.putAll(offsetWrapper.getOffsetTable());
//并持久化到从的consumerOffset.json文件中
this.brokerController.getConsumerOffsetManager().persist();
3.3、syncDelayOffset
String delayOffset = this.brokerController.getBrokerOuterAPI().getAllDelayOffset(masterAddrBak);
String fileName = StorePathConfigHelper.getDelayOffsetStorePath(this.brokerController
.getMessageStoreConfig().getStorePathRootDir());
MixAll.string2File(delayOffset, fileName);
3.4、
syncSubscriptionGroupConfig
SubscriptionGroupWrapper subscriptionWrapper =this.brokerController.getBrokerOuterAPI().getAllSubscriptionGroupConfig(masterAddrBak);
SubscriptionGroupManager subscriptionGroupManager =this.brokerController.getSubscriptionGroupManager();
subscriptionGroupManager.getDataVersion().assignNewOne(subscriptionWrapper.getDataVersion());
subscriptionGroupManager.getSubscriptionGroupTable().clear();
subscriptionGroupManager.getSubscriptionGroupTable().putAll(subscriptionWrapper.getSubscriptionGroupTable());
subscriptionGroupManager.persist();
四、思考与收获
通过上面的分享，我们基本上了解了RocketMQ的主从复制原理，其中有些思想我们可以后续借鉴下：
由于时间、精力有限，难免会有纰漏、考虑不到之处，如有问题欢迎沟通、交流。
京公网安备 11010802035086号