Spring-Cloud-Gateway 从升级到放弃
1 为什么要升级为spring-cloud-gateway？
Spring Cloud Gateway features:
-
Built on Spring Framework 5, Project Reactor and Spring Boot 2.0
-
Able to match routes on any request attribute.
-
Predicates and filters are specific to routes.
-
Hystrix Circuit Breaker integration.
-
Spring Cloud DiscoveryClient integration
-
Easy to write Predicates and Filters
-
Request Rate Limiting
-
Path Rewriting
这是官方说的，spring gateway相对spring zuul要新很多，应用也更加自由，开发体验更好。但是我在测试中发现，spring-cloud-gateway相对于zuul来说，更加出众的还是其性能，当然最后让我放弃的也是因为这一点。
网上的朋友也有做一些gateway和zuul的性能比较，大多的结论也是gateway要优于zuul，同时也更加稳定。
但是我们不能轻信，所以我也做了测试,这部分测试内容若不感兴趣可以跳过，zuul就不测试了。
2.spring-cloud-gateway的初步测试
step.1：测试准备：
1.gateway版本：2.0.1
2.服务主机：10.1.4.32，16G内存，4核虚拟机
3.测试客户端：10.1.4.34，16G内存，4核虚拟机
4.测试工具wrk
step.2：建立gateway工程并写两个测试http接口，
1.http://10.1.4.32:14077/hello [POST]
2.http://10.1.4.32:14077/test [GET]
step.3：开始测试
step.4：测试结果
[wrk@localhost wrk]$ ./wrk -t 15 -c500 -d 10 --latency http://10.1.4.32:14077/test Running 10s test @ http://10.1.4.32:14077/test 15 threads and 500 connections Thread Stats Avg Stdev Max +/- Stdev Latency 3.38ms 3.26ms 113.45ms 95.76% Req/Sec 10.84k 1.44k 26.48k 89.50% Latency Distribution 50% 2.79ms 75% 3.51ms 90% 4.21ms 99% 17.23ms 1625714 requests in 10.10s, 131.79MB read Requests/sec: 160961.07 Transfer/sec: 13.05MB
以及：
[wrk@localhost wrk]$ ./wrk -t 15 -c500 -d 10 --latency -s scripts/gateway.lua http://10.1.4.32:14077/hello Running 10s test @ http://10.1.4.32:14077/hello 15 threads and 500 connections Thread Stats Avg Stdev Max +/- Stdev Latency 5.21ms 3.96ms 255.59ms 96.75% Req/Sec 6.62k 604.79 13.72k 88.48% Latency Distribution 50% 4.78ms 75% 5.55ms 90% 6.32ms 99% 14.87ms 994374 requests in 10.10s, 539.59MB read Requests/sec: 98471.14 Transfer/sec: 53.43MB
说明，如果测试结果差别较大可能是因为测试工具的问题。
结果显示，POST方法的性能TPS达到了10W/s,而GET方法的性能TPS达到了16W/s。
这看起来很不可思议，因为正常的微服务，能达到2W/s的性能已经是良好，达到10W实在是不可思议。但是前面说了spring-cloud-gateway引入了Spring Reactor反应式编程，应对的便是这种高并发需求。
当然，即便spring-cloud-gateway给了我们很大惊喜，但是如果因此就引入了spring-cloud-gateway，那还是会有些草率，毕竟gateway是用来干什么的？是路由和过滤。继续测试。
step.5：加上路由和过滤器,在配置文件中加入下面内容
spring: cloud: gateway: routes: - id: test uri: http://10.1.4.32:14077/test predicates: - Path=/tt filters: - AddRequestParameter=foo, bar
表示，给test方法加入了路由，并且加入了官方提供的过滤器：AddRequestParameter=foo, bar
step.6：测试，并附测试结果：
[wrk@localhost wrk]$ ./wrk -t 15 -c500 -d 10 --latency http://10.1.4.32:14077/tt Running 10s test @ http://10.1.4.32:14077/tt 15 threads and 500 connections Thread Stats Avg Stdev Max +/- Stdev Latency 18.99ms 12.15ms 122.69ms 70.84% Req/Sec 1.82k 155.77 2.36k 73.94% Latency Distribution 50% 17.03ms 75% 25.49ms 90% 35.02ms 99% 57.13ms 274529 requests in 10.10s, 22.25MB read Requests/sec: 27182.88 Transfer/sec: 2.20MB
性能只剩27000/s，貌似降低了很多，但是比起zuul仍然快了不少。因为在这台机器上，测试zuul或许都不能到达2W。
那么，是不是就应该使用spring-cloud-gateway了？
3.开始使用spring-cloud-gateway
在使用上spring-cluod-gateway之后，我开始编辑自己的过滤器，需求要求写两个过滤器，修改请求体和响应体。
因为需要对特定的请求使用过滤器，所以这里使用gateway-filter，有些代码官方有，有些网友提供，两个过滤器代码大致如下：
解密过滤器，pre：
package com.newland.dc.ctid.fileter; import com.google.gson.Gson; import com.google.gson.GsonBuilder; import com.newland.dc.common.vo.RequestHeaderVo; import com.newland.dc.ctid.entity.dto.RequestDto; import io.netty.buffer.ByteBufAllocator; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.cloud.gateway.filter.GatewayFilter; import org.springframework.cloud.gateway.filter.GatewayFilterChain; import org.springframework.cloud.gateway.filter.factory.AbstractGatewayFilterFactory; import org.springframework.core.Ordered; import org.springframework.core.io.buffer.DataBuffer; import org.springframework.core.io.buffer.DataBufferUtils; import org.springframework.core.io.buffer.NettyDataBufferFactory; import org.springframework.core.style.ToStringCreator; import org.springframework.http.MediaType; import org.springframework.http.server.reactive.ServerHttpRequest; import org.springframework.http.server.reactive.ServerHttpRequestDecorator; import org.springframework.stereotype.Component; import org.springframework.web.server.ServerWebExchange; import reactor.core.publisher.Flux; import reactor.core.publisher.Mono; import java.nio.CharBuffer; import java.nio.charset.StandardCharsets; import java.util.Arrays; import java.util.List; import java.util.concurrent.atomic.AtomicReference; /** * Created by garfield on 2019/2/26. */ @Component public class DecryptGatewayFilterFactory extends AbstractGatewayFilterFactory<DecryptGatewayFilterFactory.Config>{ private static Logger log = LoggerFactory.getLogger(DecryptGatewayFilterFactory.class); public static final String DECRYPT_HEADER = "decrypt_header"; public DecryptGatewayFilterFactory() { super(Config.class); } private Gson gson = new GsonBuilder().serializeNulls().create(); @Override @SuppressWarnings("unchecked") public GatewayFilter apply(Config config) { return new DecryptGatewayFilter(config); } public class DecryptGatewayFilter implements GatewayFilter, Ordered { Config config; DecryptGatewayFilter(Config config) { this.config = config; } @Override public Mono<Void> filter(ServerWebExchange exchange, GatewayFilterChain chain) { log.debug(config.toString()); ServerHttpRequest request = exchange.getRequest(); MediaType contentType = request.getHeaders().getContentType(); boolean postRequest = "POST".equalsIgnoreCase(request.getMethodValue()) && !contentType.toString().contains("multipart/form-data"); //判断是否为POST请求 if (postRequest) { Flux<DataBuffer> body = request.getBody(); AtomicReference<String> bodyRef = new AtomicReference<>();//缓存读取的request body信息 body.subscribe(dataBuffer -> { CharBuffer charBuffer = StandardCharsets.UTF_8.decode(dataBuffer.asByteBuffer()); DataBufferUtils.release(dataBuffer); bodyRef.set(charBuffer.toString()); });//读取request body到缓存 String bodyStr = bodyRef.get();//获取request body log.debug(bodyStr);//这里是我们需要做的操作 RequestDto requestDto = gson.fromJson(bodyStr, RequestDto.class); log.debug("decrypt filter"); //save header to response header RequestHeaderVo headerVo = requestDto.getHeader(); headerVo.setAppVersion("1000"); //此处可以传递一些变量 exchange.getResponse().getHeaders().add(DECRYPT_HEADER, gson.toJson(headerVo)); DataBuffer bodyDataBuffer = stringBuffer(bodyStr); Flux<DataBuffer> bodyFlux = Flux.just(bodyDataBuffer); request = new ServerHttpRequestDecorator(request){ @Override public Flux<DataBuffer> getBody() { return bodyFlux; } };//封装我们的request } return chain.filter(exchange.mutate().request(request.mutate().header("a","200").build()).build()); }; @Override public int getOrder() { return -10; } } public static DataBuffer stringBuffer(String value) { byte[] bytes = value.getBytes(StandardCharsets.UTF_8); NettyDataBufferFactory nettyDataBufferFactory = new NettyDataBufferFactory(ByteBufAllocator.DEFAULT); DataBuffer buffer = nettyDataBufferFactory.allocateBuffer(bytes.length); buffer.write(bytes); return buffer; } @Override public ServerHttpRequest.Builder mutate(ServerHttpRequest request) { return null; } public static class Config { private boolean decrypt; public boolean isDecrypt() { return decrypt; } public void setDecrypt(boolean decrypt) { this.decrypt = decrypt; } @Override public String toString() { return new ToStringCreator(this) .append("decrypt", decrypt) .toString(); } } @Override public List<String> shortcutFieldOrder() { return Arrays.asList("decrypt"); } }
加密过滤器，使用源码的提供的修改方法，post：
package com.newland.dc.ctid.fileter; import com.google.gson.Gson; import com.newland.dc.common.vo.RequestHeaderVo; import com.newland.dc.ctid.entity.dto.RequestDto; import org.springframework.cloud.gateway.filter.GatewayFilter; import org.springframework.cloud.gateway.filter.factory.AbstractGatewayFilterFactory; import org.springframework.cloud.gateway.filter.factory.rewrite.ModifyResponseBodyGatewayFilterFactory; import org.springframework.core.style.ToStringCreator; import org.springframework.stereotype.Component; import org.springframework.web.server.ServerWebExchange; import reactor.core.publisher.Mono; import java.util.Arrays; import java.util.List; /** * @Auther: garfield * @Date: 2019/3/5 15:33 * @Description: */ @Component public class AnGatewayFilterFactory extends AbstractGatewayFilterFactory<AnGatewayFilterFactory.Config> { private Gson gson = new Gson(); public AnGatewayFilterFactory() { super(Config.class); } @Override public GatewayFilter apply(Config config) { ModifyResponseBodyGatewayFilterFactory m1 = new ModifyResponseBodyGatewayFilterFactory(null); ModifyResponseBodyGatewayFilterFactory.Config c1 = new ModifyResponseBodyGatewayFilterFactory.Config(); c1.setInClass(String.class); c1.setOutClass(String.class); c1.setNewContentType("application/json"); c1.setRewriteFunction((exchange, body) -> { ServerWebExchange ex = (ServerWebExchange) exchange; //此处更改响应体 RequestHeaderVo requestHeaderVo = new RequestHeaderVo(); RequestDto requestDto = gson.fromJson(body.toString(), RequestDto.class); requestDto.setHeader(requestHeaderVo); body = gson.toJson(requestDto); return Mono.just(body); }); return m1.apply(c1); } public static class Config { private boolean decrypt; public boolean isDecrypt() { return decrypt; } public void setDecrypt(boolean decrypt) { this.decrypt = decrypt; } @Override public String toString() { return new ToStringCreator(this) .append("encrypt", decrypt) .toString(); } } @Override public List<String> shortcutFieldOrder() { return Arrays.asList("encrypt"); } }
这里需要转移一下话题，这个过滤器修改其实有几种方法，可以自己写，也可以应用源码提供的例子。上面的两种写法已经测试都能使用，其实我还有两种方式，大同小异就是了，但也准备贴出来，也记录一下问题：
下面这个其实就是源码中的例子，只不过不引用，自己写：
@Override @SuppressWarnings("unchecked") public Mono<Void> filter(ServerWebExchange exchange, GatewayFilterChain chain) { ServerHttpResponseDecorator responseDecorator = new ServerHttpResponseDecorator(exchange.getResponse()) { @Override public Mono<Void> writeWith(Publisher<? extends DataBuffer> body) { ServerHttpRequest request = exchange.getRequest(); MediaType originalResponseContentType = exchange.getAttribute(ORIGINAL_RESPONSE_CONTENT_TYPE_ATTR); HttpHeaders httpHeaders = new HttpHeaders(); httpHeaders.setContentType(originalResponseContentType); ResponseAdapter responseAdapter = new ResponseAdapter(body, httpHeaders); DefaultClientResponse clientResponse = new DefaultClientResponse(responseAdapter, ExchangeStrategies.withDefaults()); Mono<DataBuffer> modifiedBody = clientResponse.bodyToMono(DataBuffer.class).map(encrypt(config, new RequestHeaderVo())); BodyInserter bodyInserter = BodyInserters.fromPublisher(modifiedBody, DataBuffer.class); CachedBodyOutputMessage outputMessage = new CachedBodyOutputMessage(exchange, exchange.getResponse().getHeaders()); return bodyInserter.insert(outputMessage, new BodyInserterContext()) .then(Mono.defer(() -> { Flux<DataBuffer> messageBody = outputMessage.getBody(); HttpHeaders headers = getDelegate().getHeaders(); if (headers.getContentLength() < 0 && !headers.containsKey(HttpHeaders.TRANSFER_ENCODING)) { messageBody = messageBody.doOnNext(data -> headers.setContentLength(data.readableByteCount())); } return this.getDelegate().writeWith(messageBody); })); } @Override public Mono<Void> writeAndFlushWith(Publisher<? extends Publisher<? extends DataBuffer>> body) { return writeWith(Flux.from(body) .flatMapSequential(p -> p)); } }; return chain.filter(exchange.mutate().response(responseDecorator).build()); } @Override public int getOrder() { return NettyWriteResponseFilter.WRITE_RESPONSE_FILTER_ORDER - 1; } } private Function<DataBuffer, DataBuffer> encrypt(Config config, RequestHeaderVo headerVo) { if (config.encrypt) { return (i) -> { InputStream inputStream = i.asInputStream(); byte[] bytes = new byte[0]; try { bytes = new byte[inputStream.available()]; inputStream.read(bytes); } catch (IOException e) { e.printStackTrace(); } //进行我们的操作 String body = new String(bytes); log.debug("this is response encrypt"); log.debug(body); log.debug(headerVo.toString()); // body = encryptService.responseEncrypt(body, headerVo); //进行我们的操作 return i.write(TokenGatewayFilterFactory.stringBuffer(body)); // return i.write(new String(body).getBytes()); }; } else { return i -> i; } }
这种例子中，发现修改response body的时候，会引起代码进入NioEventLoop类中的run方法，死循环无法退出，我也不清楚为什么，修改需谨慎。
另一种，跟这位网友写得差不多，只不过我没测试就是了：https://www.jianshu.com/p/9f00e0e1681c
@Override @SuppressWarnings("unchecked") public Mono<Void> filter(ServerWebExchange exchange, GatewayFilterChain chain) { return chain.filter(exchange.mutate().response(new ServerHttpResponseDecorator(exchange.getResponse()) { @Override public Mono<Void> writeWith(Publisher<? extends DataBuffer> body) { DataBufferFactory bufferFactory = exchange.getResponse().bufferFactory(); if (getStatusCode().equals(HttpStatus.OK) && body instanceof Flux) { Flux<? extends DataBuffer> fluxBody = Flux.first(body); return super.writeWith(fluxBody.map(dataBuffer -> { System.out.println(dataBuffer.readableByteCount()); byte[] content = new byte[dataBuffer.readableByteCount()]; dataBuffer.read(content); //释放掉内存 DataBufferUtils.release(dataBuffer); //responseData就是下游系统返回的内容,可以查看修改 String responseData = new String(content, Charset.forName("UTF-8")); log.debug("响应内容:{}", responseData); log.debug("this is response encrypt"); System.out.println(responseData); byte[] newContent = responseData.getBytes(); // body = encryptService.responseEncrypt(body, headerVo); byte[] uppedContent = new String(newContent, Charset.forName("UTF-8")).getBytes(); return bufferFactory.wrap(uppedContent); })); } else { log.error("响应code异常:{}", getStatusCode()); } return super.writeWith(body); } }).build()); } @Override public int getOrder() { return NettyWriteResponseFilter.WRITE_RESPONSE_FILTER_ORDER - 1; } }
这个方法会出现问题，body的截取长度经常没有完全。
我本来是到这个网址下面寻找答案，作者是这样回复的：
上面只是简单的样例，FIux是发送多个数据的，当报文长时会拆分，处理一次只能拿到一部分报文，可以使用Flux.toArray方法将数据聚合后处理，也可以参照https://www.jianshu.com/p/9b781fb1aaa0里面的响应处理。
确实是这个问题，所以我们也可以仿照他的另外一个例子写，大家可以到他的简书博客中去看，值得提醒的是，他的例子中，版本也是2.0.1,若是版本改为2.1以上，就不能用哦！
这里蛮贴一下：
package com.newland.dc.ctid.fileter; import com.google.gson.Gson; import com.newland.dc.common.vo.RequestHeaderVo; import com.newland.dc.ctid.service.SecurityService; import com.newland.dc.log.kafka.KafkaLog; import org.reactivestreams.Publisher; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.beans.factory.annotation.Value; import org.springframework.cloud.gateway.filter.GatewayFilter; import org.springframework.cloud.gateway.filter.GatewayFilterChain; import org.springframework.cloud.gateway.filter.NettyWriteResponseFilter; import org.springframework.cloud.gateway.filter.factory.AbstractGatewayFilterFactory; import org.springframework.cloud.gateway.support.*; import org.springframework.core.Ordered; import org.springframework.core.io.buffer.DataBuffer; import org.springframework.core.style.ToStringCreator; import org.springframework.http.HttpHeaders; import org.springframework.http.HttpStatus; import org.springframework.http.MediaType; import org.springframework.http.ResponseCookie; import org.springframework.http.client.reactive.ClientHttpResponse; import org.springframework.http.server.reactive.ServerHttpRequest; import org.springframework.http.server.reactive.ServerHttpRequestDecorator; import org.springframework.http.server.reactive.ServerHttpResponse; import org.springframework.http.server.reactive.ServerHttpResponseDecorator; import org.springframework.stereotype.Component; import org.springframework.util.MultiValueMap; import org.springframework.web.reactive.function.client.ExchangeStrategies; import org.springframework.web.reactive.function.server.ServerRequest; import org.springframework.web.server.ServerWebExchange; import reactor.core.publisher.Flux; import reactor.core.publisher.Mono; import java.io.IOException; import java.io.InputStream; import java.util.*; import java.util.function.BiFunction; import java.util.function.Function; import static org.springframework.cloud.gateway.support.ServerWebExchangeUtils.ORIGINAL_RESPONSE_CONTENT_TYPE_ATTR; /** * @Auther: garfield * @Date: 2019/2/28 上午10:45 * @Description: */ @Component public class EncryptGatewayFilterFactory extends AbstractGatewayFilterFactory<EncryptGatewayFilterFactory.Config> { private static Logger log = LoggerFactory.getLogger(EncryptGatewayFilterFactory.class); @Autowired private SecurityService encryptService; public EncryptGatewayFilterFactory() { super(Config.class); } // private Gson gson = new GsonBuilder().serializeNulls().create(); private Gson gson = new Gson(); @Value("${server.host:10.10.10.10}") private String serverHost; @Value("${server.port}") private String serverPort; @Override @SuppressWarnings("unchecked") public GatewayFilter apply(Config config) { return new EncryptGatewayFilter(config); } @Override public ServerHttpRequest.Builder mutate(ServerHttpRequest request) { return null; } public static class Config { private boolean encrypt; public boolean isEncrypt() { return encrypt; } public Config setEncrypt(boolean encrypt) { this.encrypt = encrypt; return this; } @Override public String toString() { return new ToStringCreator(this) .append("encrypt", encrypt) .toString(); } } @Override public List<String> shortcutFieldOrder() { return Arrays.asList("encrypt"); } public class EncryptGatewayFilter implements GatewayFilter, Ordered { Config config; EncryptGatewayFilter(Config config) { this.config = config; } @Override @SuppressWarnings("unchecked") public Mono<Void> filter(ServerWebExchange exchange, GatewayFilterChain chain) { String trace = exchange.getRequest().getHeaders().getFirst("trace"); ServerRequest serverRequest = new DefaultServerRequest(exchange); return serverRequest.bodyToMono(String.class).flatMap(reqBody -> { //重写原始请求 ServerHttpRequestDecorator decorator = new ServerHttpRequestDecorator(exchange.getRequest()) { @Override public HttpHeaders getHeaders() { HttpHeaders httpHeaders = new HttpHeaders(); httpHeaders.putAll(super.getHeaders()); return httpHeaders; } @Override public Flux<DataBuffer> getBody() { //打印原始请求日志 log.info("[Trace:{}]-gateway request:headers=[{}],body=[{}]", trace, getHeaders(), reqBody); return Flux.just(reqBody).map(bx -> exchange.getResponse().bufferFactory().wrap(bx.getBytes())); } }; //重写原始响应 BodyHandlerServerHttpResponseDecorator responseDecorator = new BodyHandlerServerHttpResponseDecorator( initBodyHandler(exchange), exchange.getResponse()); return chain.filter(exchange.mutate().request(decorator).response(responseDecorator).build()); }); } @Override public int getOrder() { return NettyWriteResponseFilter.WRITE_RESPONSE_FILTER_ORDER - 1; } } public interface BodyHandlerFunction extends BiFunction<ServerHttpResponse, Publisher<? extends DataBuffer>, Mono<Void>> { } protected BodyHandlerFunction initBodyHandler(ServerWebExchange exchange) { return (resp, body) -> { //拦截 MediaType originalResponseContentType = exchange.getAttribute(ORIGINAL_RESPONSE_CONTENT_TYPE_ATTR); HttpHeaders httpHeaders = new HttpHeaders(); httpHeaders.setContentType(originalResponseContentType); DefaultClientResponseAdapter clientResponseAdapter = new DefaultClientResponseAdapter(body, httpHeaders); Mono<String> bodyMono = clientResponseAdapter.bodyToMono(String.class); //此处可以获得前面放置的参数 return bodyMono.flatMap((respBody) -> { // 打印返回响应日志 System.out.println(respBody); return resp.writeWith(Flux.just(respBody).map(bx -> resp.bufferFactory().wrap(bx.getBytes()))); }).then(); }; } public static class DefaultClientResponseAdapter extends DefaultClientResponse { /** * @param body * @param httpHeaders */ public DefaultClientResponseAdapter(Publisher<? extends DataBuffer> body, HttpHeaders httpHeaders) { this(new ResponseAdapter(body, httpHeaders), ExchangeStrategies.withDefaults()); } /** * @param response * @param strategies */ public DefaultClientResponseAdapter(ClientHttpResponse response, ExchangeStrategies strategies) { super(response, strategies); } /** * ClientHttpResponse 适配器 */ static class ResponseAdapter implements ClientHttpResponse { /** * 响应数据 */ private final Flux<DataBuffer> flux; /** * */ private final HttpHeaders headers; public ResponseAdapter(Publisher<? extends DataBuffer> body, HttpHeaders headers) { this.headers = headers; if (body instanceof Flux) { flux = (Flux) body; } else { flux = ((Mono) body).flux(); } } @Override public Flux<DataBuffer> getBody() { return flux; } @Override public HttpHeaders getHeaders() { return headers; } @Override public HttpStatus getStatusCode() { return null; } @Override public int getRawStatusCode() { return 0; } @Override public MultiValueMap<String, ResponseCookie> getCookies() { return null; } } } class BodyHandlerServerHttpResponseDecorator extends ServerHttpResponseDecorator { /** * body 处理拦截器 */ private BodyHandlerFunction bodyHandler = initDefaultBodyHandler(); /** * 构造函数 * * @param bodyHandler * @param delegate */ public BodyHandlerServerHttpResponseDecorator(BodyHandlerFunction bodyHandler, ServerHttpResponse delegate) { super(delegate); if (bodyHandler != null) { this.bodyHandler = bodyHandler; } } @Override public Mono<Void> writeWith(Publisher<? extends DataBuffer> body) { //body 拦截处理器处理响应 return bodyHandler.apply(getDelegate(), body); } @Override public Mono<Void> writeAndFlushWith(Publisher<? extends Publisher<? extends DataBuffer>> body) { return writeWith(Flux.from(body).flatMapSequential(p -> p)); } /** * 默认body拦截处理器 * * @return */ private BodyHandlerFunction initDefaultBodyHandler() { return (resp, body) -> resp.writeWith(body); } } }
那么万事具备，代码都写好了，我们又需要进行性能测试。这边要记住，我用的是官方的那个例子，其他的写法也用过了，但是结果差不多。
4.再一次测试spring-cloud-gateway 网关路由性能
step.1：性能测试，改一下配置，表示加入了过滤器。这里为什么只有一个过滤器，因为这个过滤器问题比较大，过程就略过了。
- id: an uri: http://10.1.4.32:14077/hello predicates: - Path=/an filters: - An
经过多次测试，其他的过滤器都还好，只有修改response body的过滤器，严重影响性能，且有读写错误。
step.2：测试，以及测试结果
[wrk@localhost wrk]$ ./wrk -t 15 -c500 -d 10 --latency -s scripts/gateway.lua http://10.1.4.32:14077/an Running 10s test @ http://10.1.4.32:14077/an 15 threads and 500 connections Thread Stats Avg Stdev Max +/- Stdev Latency 1.03s 488.75ms 2.00s 60.62% Req/Sec 26.59 13.84 80.00 67.60% Latency Distribution 50% 931.54ms 75% 1.45s 90% 1.76s 99% 1.97s 3848 requests in 10.10s, 1.64MB read Socket errors: connect 0, read 0, write 0, timeout 458 Requests/sec: 381.05 Transfer/sec: 166.71KB
结果多出一行，socket错误，而且还是超时,而且，日志中也存在错误：
2019-03-06T16:09:33,396|INFO ||AsyncResolver-bootstrap-executor-0||||Resolving eureka endpoints via configuration 2019-03-06T16:10:38,268|ERROR||reactor-http-server-epoll-18||||Unhandled failure: Connection has been closed, response already set (status=200) 2019-03-06T16:10:38,268|WARN ||reactor-http-server-epoll-18||||Handling completed with error: Connection has been closed 2019-03-06T16:10:38,269|ERROR||reactor-http-server-epoll-18||||Unhandled failure: null, response already set (status=200) 2019-03-06T16:10:38,269|WARN ||reactor-http-server-epoll-18||||Handling completed with error: null 2019-03-06T16:10:38,294|ERROR||reactor-http-server-epoll-18||||Unhandled failure: syscall:write(..) failed: 断开的管道, response already set (status=null) 2019-03-06T16:10:38,294|WARN ||reactor-http-server-epoll-18||||Handling completed with error: syscall:write(..) failed: 断开的管道 2019-03-06T16:10:38,306|ERROR||reactor-http-server-epoll-23||||Unhandled failure: syscall:write(..) failed: 断开的管道, response already set (status=null) 2019-03-06T16:10:38,306|WARN ||reactor-http-server-epoll-23||||Handling completed with error: syscall:write(..) failed: 断开的管道 2019-03-06T16:14:33,397|INFO ||AsyncResolver-bootstrap-executor-0||||Resolving eureka endpoints via configuration
这个问题很严重了，因为单个请求的时候，并不会报错，这个错误只发生在高并发压测下，无法追踪。最重要的是，我们看到性能只剩下300/s，这是万万不能接受的，生产更不能接收。
这个问题很难解释，因为我们采用的是官方提供的写法，我们回头看官方的修改response 类，好吧，不用看了，因为：
package org.springframework.cloud.gateway.filter.factory.rewrite; /** * This filter is BETA and may be subject to change in a future release. */ public class ModifyResponseBodyGatewayFilterFactory extends AbstractGatewayFilterFactory<ModifyResponseBodyGatewayFilterFactory.Config> {
官方已经说了，这是测试版本，不顶用。
不死心，又想起了gateway提供的GlobalFilter，将刚才的代码写到全局过滤器中再试试，但是结果相同！
凉凉...
跪求结论跟我不同的启发文档，或者只能等下一版本了。
