1、HTTP长连接
HTTP长连接，也称为持久连接，是一种使用同一个TCP连接来发送和接收多个HTTP请求/应答的方法，而不是为每一个新的请求/应答打开新的TCP连接。这种方式由于通信连接一直存在，因此可以减少建立和关闭连接的开销，提高通信效率。因为HTTP长连接的本质就是保持TCP的连接在每次请求响应之后不断开，与其说是HTTP长连接，不如说是TCP的长连接。
那么tomcat作为最常用的WEB容器，是怎么处理HTTP的长连接呢？
2、tomcat处理长连接
在tomcat的Poller线程中，监听已连接套接字以保持连接，并轮询以检查数据是否可用。具体来说，Poller线程使用NIO架构，通过内部的Selector对象向内核查询Channel的状态，一旦发现可读事件，就会生成任务类SocketProcessor，并将其交给Executor去处理。
public void run() {
// Loop until destroy() is called
while (true) {
boolean hasEvents = false;
try {
if (!close) {
hasEvents = events();
if (wakeupCounter.getAndSet(-1) > 0) {
// If we are here, means we have other stuff to do
// Do a non blocking select
keyCount = selector.selectNow();
} else {
keyCount = selector.select(selectorTimeout);
}
wakeupCounter.set(0);
}
if (close) {
events();
timeout(0, false);
try {
selector.close();
} catch (IOException ioe) {
log.error(sm.getString("endpoint.nio.selectorCloseFail"), ioe);
}
break;
}
// Either we timed out or we woke up, process events first
if (keyCount == 0) {
hasEvents = (hasEvents | events());
}
} catch (Throwable x) {
ExceptionUtils.handleThrowable(x);
log.error(sm.getString("endpoint.nio.selectorLoopError"), x);
continue;
}
Iterator<SelectionKey> iterator =
keyCount > 0 ? selector.selectedKeys().iterator() : null;
// Walk through the collection of ready keys and dispatch
// any active event.
while (iterator != null && iterator.hasNext()) {
SelectionKey sk = iterator.next();
iterator.remove();
NioSocketWrapper socketWrapper = (NioSocketWrapper) sk.attachment();
// Attachment may be null if another thread has called
// cancelledKey()
if (socketWrapper != null) {
processKey(sk, socketWrapper);
}
}
// Process timeouts
timeout(keyCount,hasEvents);
}
getStopLatch().countDown();
}
Poller线程的run方法是while(true)死循环，主要监听注册的socket上是否有已就绪事件，如果有的话就调用processKey(sk, socketWrapper)方法交由线程池处理，最后调用了timeout方法。
protected void timeout(int keyCount, boolean hasEvents) {
long now = System.currentTimeMillis();
// nextExpiration初始化是0
if (nextExpiration > 0 && (keyCount > 0 || hasEvents) && (now < nextExpiration) && !close) {
return;
}
int keycount = 0;
try {
// 遍历注册到selector上所有的socket
for (SelectionKey key : selector.keys()) {
keycount++;
NioSocketWrapper socketWrapper = (NioSocketWrapper) key.attachment();
try {
if (socketWrapper == null) {
// We don't support any keys without attachments
if (key.isValid()) {
key.cancel();
}
} else if (close) {
key.interestOps(0);
// Avoid duplicate stop calls
socketWrapper.interestOps(0);
socketWrapper.close();
// 如果注册的事件是读写事件
} else if (socketWrapper.interestOpsHas(SelectionKey.OP_READ) ||
socketWrapper.interestOpsHas(SelectionKey.OP_WRITE)) {
boolean readTimeout = false;
boolean writeTimeout = false;
// 检查读超时
if (socketWrapper.interestOpsHas(SelectionKey.OP_READ)) {
// 用当前时间-上次读时间
long delta = now - socketWrapper.getLastRead();
long timeout = socketWrapper.getReadTimeout();
if (timeout > 0 && delta > timeout) {
readTimeout = true;
}
}
// Check for write timeout
if (!readTimeout && socketWrapper.interestOpsHas(SelectionKey.OP_WRITE)) {
long delta = now - socketWrapper.getLastWrite();
long timeout = socketWrapper.getWriteTimeout();
if (timeout > 0 && delta > timeout) {
writeTimeout = true;
}
}
// 如果已经超时
if (readTimeout || writeTimeout) {
key.interestOps(0);
// Avoid duplicate timeout calls
socketWrapper.interestOps(0);
socketWrapper.setError(new SocketTimeoutException());
if (readTimeout && socketWrapper.readOperation != null) {
if (!socketWrapper.readOperation.process()) {
socketWrapper.close();
}
} else if (writeTimeout && socketWrapper.writeOperation != null) {
if (!socketWrapper.writeOperation.process()) {
socketWrapper.close();
}
// processSocket中对将socket进行关闭
} else if (!processSocket(socketWrapper, SocketEvent.ERROR, true)) {
socketWrapper.close();
}
}
}
} catch (CancelledKeyException ckx) {
if (socketWrapper != null) {
socketWrapper.close();
}
}
}
} catch (ConcurrentModificationException cme) {
// See https://bz.apache.org/bugzilla/show_bug.cgi?id=57943
log.warn(sm.getString("endpoint.nio.timeoutCme"), cme);
}
// For logging purposes only
long prevExp = nextExpiration;
// nextExpiration重新赋值 当前时间+1s，socketProperties.getTimeoutInterval()默认1000
nextExpiration = System.currentTimeMillis() +
socketProperties.getTimeoutInterval();
if (log.isTraceEnabled()) {
log.trace("timeout completed: keys processed=" + keycount +
"; now=" + now + "; nextExpiration=" + prevExp +
"; keyCount=" + keyCount + "; hasEvents=" + hasEvents +
"; eval=" + ((now < prevExp) && (keyCount>0 || hasEvents) && (!close) ));
}
}
timeout方法主要做了以下事：
- 判断是否要进行轮询所有socket进行超时判断
- 遍历所有socket，拿到上次读写的事件，与当前时间对比，是否已超时
- 如果已超时，对相关socket进行关闭处理
- 重置nextExpiration值，默认每秒都会对所有socket进行超时轮询判断
在进行对socket读取时会把keepAliveTimeout参数赋值给ReadTimeout(前提，开启长连接，tomcat已经默认开启长连接)
if (keptAlive) {
// Haven't read any request data yet so use the keep-alive
// timeout.
wrapper.setReadTimeout(keepAliveTimeout);
}
每次对socket进行读取后，也会调用updateLastRead方法更新上次读取时间
if (to.remaining() >= limit) {
to.limit(to.position() + limit);
nRead = fillReadBuffer(block, to);
if (log.isDebugEnabled()) {
log.debug("Socket: [" + this + "], Read direct from socket: [" + nRead + "]");
}
updateLastRead();
}
3、总结
tomcat处理Http长连接是在Poller线程中的timeout方法，最长每秒都会对所有的socket进行遍历，上次读写数据的时间与当前时间和参数配置的keep-alive-timeout时间进行判断是否已经超时（前提开启长连接），如果已经超时则对相应的socket进行关闭
