继续浏览精彩内容
慕课网APP
程序员的梦工厂
打开
继续
感谢您的支持,我会继续努力的
赞赏金额会直接到老师账户
将二维码发送给自己后长按识别
微信支付
支付宝支付

Spring-Cloud-Gateway 从升级到放弃

胡说叔叔
关注TA
已关注
手记 495
粉丝 130
获赞 581

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   

复制代码

[wrk@localhost wrk]$ ./wrk  -t 15 -c500 -d 10 --latency  http://10.1.4.32:14077/testRunning 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: 160961.07Transfer/sec:     13.05MB

复制代码

以及:

复制代码

[wrk@localhost wrk]$ ./wrk  -t 15 -c500 -d 10 --latency -s scripts/gateway.lua  http://10.1.4.32:14077/helloRunning 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:  98471.14Transfer/sec:     53.43MB

复制代码

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

复制代码

spring:
  cloud:
    gateway:
      routes:      - id: test
        uri: http://10.1.4.32:14077/test        predicates:        - Path=/tt
        filters:        - AddRequestParameter=foo, bar

复制代码

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:

复制代码

[wrk@localhost wrk]$ ./wrk  -t 15 -c500 -d 10 --latency  http://10.1.4.32:14077/ttRunning 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:  27182.88Transfer/sec:      2.20MB

复制代码

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");
    }
}

View Code

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");
    }
}

View Code

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;
        }

    }

View Code

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;
        }
    }

View Code

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);
        }
    }
}

View Code

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

复制代码

[wrk@localhost wrk]$ ./wrk  -t 15 -c500 -d 10 --latency -s scripts/gateway.lua  http://10.1.4.32:14077/anRunning 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 458Requests/sec:    381.05Transfer/sec:    166.71KB

复制代码

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:

复制代码

2019-03-06T16:09:33,396|INFO ||AsyncResolver-bootstrap-executor-0||||Resolving eureka endpoints via configuration2019-03-06T16:10:38,268|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 closed2019-03-06T16:10:38,269|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: null2019-03-06T16:10:38,294|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

复制代码

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。

作者:夏虫而已

原文出处:https://www.cnblogs.com/garfieldcgf/p/10484119.html  

打开App,阅读手记
0人推荐
发表评论
随时随地看视频慕课网APP