使用 WebLogic 12.2.1 处理并发 JAX-RS 请求

Question

我编写了一个在 WebLogic 12.2.1 上运行的 JAX-RS Web 服务方法，以测试它可以处理多少并发请求。我故意让这个方法需要 5 分钟来执行。

@Singleton
@Path("Services")
@ApplicationPath("resources")
public class Services extends Application {
    private static int count = 0;

    private static synchronized int addCount(int a) {
        count = count + a;
        return count;
    }

    @GET
    @Path("Ping")
    public Response ping(@Context HttpServletRequest request) {
        int c = addCount(1);
        logger.log(INFO, "Method entered, total running requests: [{0}]", c);
        try {
            Thread.sleep(300000);
        } catch (InterruptedException exception) {
        }
        c = addCount(-1);
        logger.log(INFO, "Exiting method, total running requests: [{0}]", c);
        return Response.ok().build();
    }
}

我还编写了一个独立的客户端程序来向该服务发送 500 个并发请求。客户端为每个请求使用一个线程。

据我了解，WebLogic默认最多有400个线程，也就是说它可以同时处理400个请求。这个数字在下面我的测试结果中得到了证实。如您所见，在前 5 分钟内，从 10:46:31 开始，仅处理了 400 个请求。

23/08/2016 10:46:31.393 [132] [INFO] [Services.ping] - Method entered, total running requests: [1]
23/08/2016 10:46:31.471 [204] [INFO] [Services.ping] - Method entered, total running requests: [2]
23/08/2016 10:46:31.471 [66] [INFO] [Services.ping] - Method entered, total running requests: [3]
23/08/2016 10:46:31.471 [210] [INFO] [Services.ping] - Method entered, total running requests: [4]
23/08/2016 10:46:31.471 [206] [INFO] [Services.ping] - Method entered, total running requests: [5]
23/08/2016 10:46:31.487 [207] [INFO] [Services.ping] - Method entered, total running requests: [6]
23/08/2016 10:46:31.487 [211] [INFO] [Services.ping] - Method entered, total running requests: [7]
23/08/2016 10:46:31.487 [267] [INFO] [Services.ping] - Method entered, total running requests: [8]
23/08/2016 10:46:31.487 [131] [INFO] [Services.ping] - Method entered, total running requests: [9]
23/08/2016 10:46:31.502 [65] [INFO] [Services.ping] - Method entered, total running requests: [10]
23/08/2016 10:46:31.518 [265] [INFO] [Services.ping] - Method entered, total running requests: [11]
23/08/2016 10:46:31.565 [266] [INFO] [Services.ping] - Method entered, total running requests: [12]
23/08/2016 10:46:35.690 [215] [INFO] [Services.ping] - Method entered, total running requests: [13]
23/08/2016 10:46:35.690 [269] [INFO] [Services.ping] - Method entered, total running requests: [14]
23/08/2016 10:46:35.690 [268] [INFO] [Services.ping] - Method entered, total running requests: [15]
23/08/2016 10:46:35.690 [214] [INFO] [Services.ping] - Method entered, total running requests: [16]
23/08/2016 10:46:35.690 [80] [INFO] [Services.ping] - Method entered, total running requests: [17]
23/08/2016 10:46:35.690 [79] [INFO] [Services.ping] - Method entered, total running requests: [18]
23/08/2016 10:46:35.690 [152] [INFO] [Services.ping] - Method entered, total running requests: [19]
23/08/2016 10:46:37.674 [158] [INFO] [Services.ping] - Method entered, total running requests: [20]
23/08/2016 10:46:37.674 [155] [INFO] [Services.ping] - Method entered, total running requests: [21]
23/08/2016 10:46:39.674 [163] [INFO] [Services.ping] - Method entered, total running requests: [22]
23/08/2016 10:46:39.705 [165] [INFO] [Services.ping] - Method entered, total running requests: [23]
23/08/2016 10:46:39.705 [82] [INFO] [Services.ping] - Method entered, total running requests: [24]
23/08/2016 10:46:39.705 [166] [INFO] [Services.ping] - Method entered, total running requests: [25]
23/08/2016 10:46:41.690 [84] [INFO] [Services.ping] - Method entered, total running requests: [26]
23/08/2016 10:46:41.690 [160] [INFO] [Services.ping] - Method entered, total running requests: [27]
23/08/2016 10:46:43.690 [226] [INFO] [Services.ping] - Method entered, total running requests: [28]
23/08/2016 10:46:43.705 [162] [INFO] [Services.ping] - Method entered, total running requests: [29]
....
....
23/08/2016 10:50:52.008 [445] [INFO] [Services.ping] - Method entered, total running requests: [398]
23/08/2016 10:50:52.008 [446] [INFO] [Services.ping] - Method entered, total running requests: [399]
23/08/2016 10:50:54.008 [447] [INFO] [Services.ping] - Method entered, total running requests: [400]
23/08/2016 10:51:31.397 [132] [INFO] [Services.ping] - Exiting method, total running requests: [399]
23/08/2016 10:51:31.475 [207] [INFO] [Services.ping] - Exiting method, total running requests: [398]
23/08/2016 10:51:31.475 [207] [INFO] [Services.ping] - Method entered, total running requests: [399]
....
....

但我不明白的是，为什么前 400 个请求没有被 service 方法同时处理？从测试结果可以看出，第一个请求是在 10:46:31.393 处理的，而第 400 个请求是在 10:50:54.008 处理的，比 4 分钟多了。

如果我们查看 access.log，我们可以看到 WebLogic 在 10:46:31 到 10:46:35 之间收到了所有 500 个请求。所以看起来即使WebLogic在很短的时间内收到请求，它也没有分配线程和调用服务方法那么快。

10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:31 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
....
....
10.204.133.176 - - [23/Aug/2016:10:46:35 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:35 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:35 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 
10.204.133.176 - - [23/Aug/2016:10:46:35 +0800] "GET /Test/Services/Ping HTTP/1.1" 200 0 

已编辑

添加了工作管理器以定义至少 400 个线程。

weblogic.xml

<wls:work-manager>
    <wls:name>HighPriorityWorkManager</wls:name>
    <wls:fair-share-request-class>
        <wls:name>HighPriority</wls:name> 
        <wls:fair-share>100</wls:fair-share> 
    </wls:fair-share-request-class>
    <wls:min-threads-constraint>
        <wls:name>MinThreadsCount</wls:name>
        <wls:count>400</wls:count>
    </wls:min-threads-constraint>
</wls:work-manager>

web.xml

     <init-param>
        <param-name>wl-dispatch-policy</param-name>
        <param-value>HighPriorityWorkManager</param-value>
     </init-param>

Answer 1

这就是 weblogic 扩展线程池的方式（它们是“自我调整”），它不会立即启动 400 个线程。更多的是缓慢增加线程以最大化吞吐量。

https://docs.oracle.com/cd/E24329_01/web.1211/e24432/self_tuned.htm#CNFGD113