原文地址
http://my.oschina.net/momohuang/blog/114552
先说一下,本来是想自己写socket ,启动一个简单点的web服务用于接收数据的。写完之后,发现会有各种情况没有考虑到的,很有可能出现问题,而且,太折腾了。于是,就用了netty去其web服务,另外,我也觉得netty基本上是最简单的web服务了吧,如果童鞋们有其他推荐的话,就留个言呗。
1、server
public class AdminServer {
protected static final Log log = LogFactory.getLog(AdminServer.class);
public static void main(String[] args) {
log.info("start app");
start(8088);
// System.out.println("admin start on "+1);
}
public static void start(int port) {
// 配置服务器-使用java线程池作为解释线程
ServerBootstrap bootstrap = new ServerBootstrap(new NioServerSocketChannelFactory(Executors.newCachedThreadPool(), Executors.newCachedThreadPool()));
// 设置 pipeline factory.
bootstrap.setPipelineFactory(new ServerPipelineFactory());
// 绑定端口
bootstrap.bind(new InetSocketAddress(port));
System.out.println("admin start on "+port);
ServiceLocator.initServiceLocator();
}
private static class ServerPipelineFactory implements
ChannelPipelineFactory {
public ChannelPipeline getPipeline() throws Exception {
// Create a default pipeline implementation.
ChannelPipeline pipeline = Channels.pipeline();
pipeline.addLast("decoder", new HttpRequestDecoder());
pipeline.addLast("encoder", new HttpResponseEncoder());
//http处理handler
pipeline.addLast("handler", new AdminServerHandler());
return pipeline;
}
}
}
启动了服务,绑定了8088端口。
2、当客户端给服务端post数据的时候,如果数据超过50K,这个时候服务端接受到的post内容是空的了。这是因为超过了 服务端默认的post的长度的最大值。
http协议里边,本来是没有对post的长度进行限制,但是,无论是系统层面或者是服务端层面的,都会对post的长度进行限制,这个也有利于网络安全。
3、在netty中的解决方法
private static class ServerPipelineFactory implements
ChannelPipelineFactory {
public ChannelPipeline getPipeline() throws Exception {
// Create a default pipeline implementation.
ChannelPipeline pipeline = Channels.pipeline();
// pipeline.addFirst("frameDecoder", new LengthFieldBasedFrameDecoder(100000000,0,4,0,4));
pipeline.addLast("decoder", new HttpRequestDecoder());
pipeline.addLast("encoder", new HttpResponseEncoder());
// pipeline.addLast("streamer", new ChunkedWriteHandler());
pipeline.addLast("aggregator", new HttpChunkAggregator(65536));//设置块的最大字节数
//http处理handler
pipeline.addLast("handler", new AdminServerHandler());
return pipeline;
}
}
加上
pipeline.addLast("aggregator", new HttpChunkAggregator(65536))
之后,设置默认的chunk最大为 65536,这样,就可以接受最大post的内容大小为 65536。
这样有一个不好的地方,就是这个大小不好控制,开大了,会浪费空间。并且在接受到的字符串的最后,会出现空白的字符串,这是由于post的内容长度小于chunk里边的ChannelBuffer的数组的大小,程序给予补全。
4、自己设置,自己读取chunk
加上
pipeline.addLast("streamer", new ChunkedWriteHandler());
设置为给位 分开一个个chunk去接受信息。
public boolean excuteChunk(ChannelHandlerContext ctx, MessageEvent e)
throws TooLongFrameException {
// HttpMessage currentMessage = e.getMessage();
if (e.getMessage() instanceof HttpMessage) {
HttpMessage m = (HttpMessage) e.getMessage();
if (m.isChunked()) {
// A chunked message - remove 'Transfer-Encoding' header,
// initialize the cumulative buffer, and wait for incoming
// chunks.
List<String> encodings = m
.getHeaders(HttpHeaders.Names.TRANSFER_ENCODING);
encodings.remove(HttpHeaders.Values.CHUNKED);
if (encodings.isEmpty()) {
m.removeHeader(HttpHeaders.Names.TRANSFER_ENCODING);
}
m.setContent(ChannelBuffers.dynamicBuffer(e.getChannel()
.getConfig().getBufferFactory()));
this.currentMessage = m;
} else {
// Not a chunked message - pass through.
this.currentMessage = null;
}
return false;
} else if (e.getMessage() instanceof HttpChunk) {
// Sanity check
if (currentMessage == null) {
throw new IllegalStateException("received "
+ HttpChunk.class.getSimpleName() + " without "
+ HttpMessage.class.getSimpleName());
}
// Merge the received chunk into the content of the current message.
HttpChunk chunk = (HttpChunk) e.getMessage();
ChannelBuffer content = currentMessage.getContent();
if (content.readableBytes() > maxContentLength
- chunk.getContent().readableBytes()) {
throw new TooLongFrameException("HTTP content length exceeded "
+ maxContentLength + " bytes.");
}
content.writeBytes(chunk.getContent());
if (chunk.isLast()) {
this.currentMessage = null;
currentMessage.setHeader(HttpHeaders.Names.CONTENT_LENGTH