Hystrix：Apache http客户端请求不中断

Question

我有一个 apache HTTP 客户端定义如下：

private static HttpClient httpClient = null;
HttpParams httpParams = new BasicHttpParams();
httpParams.setParameter(CoreProtocolPNames.USE_EXPECT_CONTINUE, Boolean.TRUE);
httpParams.setParameter(CoreProtocolPNames.USER_AGENT, "ABC");

HttpConnectionParams.setStaleCheckingEnabled(httpParams, Boolean.TRUE);

SSLSocketFactory sf = SSLSocketFactory.getSocketFactory();

SchemeRegistry schemeRegistry = new SchemeRegistry();
schemeRegistry.register(new Scheme("http", 80, PlainSocketFactory.getSocketFactory()));
schemeRegistry.register(new Scheme("https", 443, sf));

//Initialize the http connection pooling
PoolingClientConnectionManager connectionManager = new PoolingClientConnectionManager(schemeRegistry);

// Initialize the connection parameters for performance tuning
connectionManager.setMaxTotal(12);
connectionManager.setDefaultMaxPerRoute(10);

httpClient = new DefaultHttpClient(connectionManager, httpParams);

我有一个 hystrix 命令 play 并启用了以下属性：

hystrix.command.play.execution.isolation.thread.timeoutInMilliseconds=1
hystrix.command.play.execution.isolation.thread.interruptOnTimeout=true

命令本身定义如下：

    @HystrixCommand(groupKey="play_group",commandKey="play")
    public String process(String request) throws UnsupportedOperationException, IOException, InterruptedException {
        System.out.println("Before -  process method : " + request);
        callHttpClient(request);
        System.out.println("After -  process method" + request);
        return "";
    }

    private void callHttpClient(String request) throws ClientProtocolException, IOException, InterruptedException {
        HttpGet get = new HttpGet("http://www.google.co.in");
        HttpResponse response = httpClient.execute(get);
        System.out.println("Response:" + response);
    }

我现在尝试循环执行命令 5 次：

    public static void main(String[] args) throws UnsupportedOperationException, IOException, InterruptedException {
        ApplicationContext ctx = new ClassPathXmlApplicationContext("applicationContextTest.xml");
        HystrixPlayground obj = ctx.getBean(HystrixPlayground.class);

        long t1 = System.currentTimeMillis();
        for (int i = 0; i < 5; i++) {
            try{
                System.out.println(obj.process("test" + i));
            } catch(Exception ex) {
                System.out.println(ex);
            }
            long t2 = System.currentTimeMillis();
            System.out.println("Time(ms) : ---->" + (t2 - t1));

超时设置为 1 毫秒，因此 process 方法抛出 HystrixRunTimeException。但是，http 请求继续执行并打印 "After - process method" 字符串。

我只在 http 客户端请求中看到过这种行为。如果 http 请求被诸如线程睡眠或非常大的 for 循环之类的任何其他内容替换，hystrix 线程将按预期中断。

有没有人知道为什么会发生这种情况？

Answer 1

原因是在 Java 中中断线程不会 "force stop" 它。相反，调用 Thread.interrupt() 只是设置一个标志，该标志可以但不必由运行线程解释。在此处查看更多信息：What does java.lang.Thread.interrupt() do?

Apache HTTP 客户端不解释这个标志（因为 JRE 的套接字读取操作不能被中断）。因此，HTTP 请求没有被取消，只是完成了。

Answer 2

似乎在过去，一些 IO 操作，在某些平台（solaris）上可以使用 Thread.interrupt() 中断。现在对于 java.io， 似乎不再是 possible/specified，但对于 java.nio 仍然是可能的。（有关详细信息，请参阅 Java bug 4385444 and Java bug 7188233）

所以即使 with some platforms/versions it seems still possible 使用 Thread.interrupt() 中断阻塞 IO，这似乎显然不推荐：

https://www.oreilly.com/library/view/java-threads-second/1565924185/ch04s05.html#ch04-8-fm2xml

https://wiki.sei.cmu.edu/confluence/display/java/THI04-J.+Ensure+that+threads+performing+blocking+operations+can+be+terminated

所以httpclient没有任何问题。使用 Hystrix，您应该 abort() 请求 Hystrix 超时。如果你使用fallback，你可以在getFallback()中调用它，如果没有你可以在得到结果后调用它。例如：

try { hystrixCommand.execute(); } catch (HystrixRuntimeException e) { switch (e.getFailureType()) { case TIMEOUT: // supposing you add a getHttpRequest getter on your command. hystrixCommand.getHttpRequest().abort(); } }

Hystrix：Apache http客户端请求不中断

Hystrix: Apache http client requests are not interrupted

java

timeout

hystrix