EagerThreadPoolExecutor 在Jmeter压测时出现线程池耗尽问题

[workingonescape]

最近在学习Dubbo源码，当学习到Dubbo中有关EagerThreadPool相关知识时遇到了一个问题，下面是我的dubbo-demo的一些配置：

然后我使用Jmeter进行压测，模拟高并发场景

紧接着，使用 jmeter -n -t命令执行压测

在这个过程中，会出现 Dubbo线程池耗尽的警告信息

[08/09/22 10:13:07:886 CST] NettyServerWorker-3-4  WARN support.AbortPolicyWithReport:  [DUBBO] Thread pool is EXHAUSTED! Thread Name: dubbo-demo-annotation-provider-198.18.0.1:20880, Pool Size: 10 (active: 1, core: 5, max: 10, largest: 10), Task: 92341 (completed: 92340), Executor status:(isShutdown:false, isTerminated:false, isTerminating:false), in dubbo://198.18.0.1:20880!, dubbo version: , current host: 198.18.0.1
[08/09/22 10:13:07:886 CST] NettyServerWorker-3-5  WARN support.AbortPolicyWithReport:  [DUBBO] Thread pool is EXHAUSTED! Thread Name: dubbo-demo-annotation-provider-198.18.0.1:20880, Pool Size: 10 (active: 0, core: 5, max: 10, largest: 10), Task: 92342 (completed: 92341), Executor status:(isShutdown:false, isTerminated:false, isTerminating:false), in dubbo://198.18.0.1:20880!, dubbo version: , current host: 198.18.0.1
[08/09/22 10:13:07:886 CST] NettyServerWorker-3-8  WARN support.AbortPolicyWithReport:  [DUBBO] Thread pool is EXHAUSTED! Thread Name: dubbo-demo-annotation-provider-198.18.0.1:20880, Pool Size: 10 (active: 0, core: 5, max: 10, largest: 10), Task: 92342 (completed: 92341), Executor status:(isShutdown:false, isTerminated:false, isTerminating:false), in dubbo://198.18.0.1:20880!, dubbo version: , current host: 198.18.0.1
[08/09/22 10:13:07:886 CST] NettyServerWorker-3-9  WARN support.AbortPolicyWithReport:  [DUBBO] Thread pool is EXHAUSTED! Thread Name: dubbo-demo-annotation-provider-198.18.0.1:20880, Pool Size: 10 (active: 0, core: 5, max: 10, largest: 10), Task: 92342 (completed: 92341), Executor status:(isShutdown:false, isTerminated:false, isTerminating:false), in dubbo://198.18.0.1:20880!, dubbo version: , current host: 198.18.0.1
[08/09/22 10:13:07:886 CST] NettyServerWorker-3-6  WARN support.AbortPolicyWithReport:  [DUBBO] Thread pool is EXHAUSTED! Thread Name: dubbo-demo-annotation-provider-198.18.0.1:20880, Pool Size: 10 (active: 0, core: 5, max: 10, largest: 10), Task: 92341 (completed: 92340), Executor status:(isShutdown:false, isTerminated:false, isTerminating:false), in dubbo://198.18.0.1:20880!, dubbo version: , current host: 198.18.0.1

本来也很正常，Jmeter文件里设置了线程数为 100

但是报错信息的 Pool的activeCount竟然不是10，如果线程池耗尽的话，这里的activeCount应该是10才对。

这是为什么？

[workingonescape]

然后我改造了一下，增加了一下日志打印参数，如下：

之后还是使用Jmeter进行压测，结果如下：

submittedTaskCount、active和Pool Size还是对不上

[workingonescape]

这是日志 provider.log

[AlbumenJ]

debug 看了下，active 小于 pool 的原因是中间的 queue 存在竞争以及 worker 获取任务不够及时导致的。queue 队列长度只有 1，在高并发场景下会导致 worker 来不及获取任务就有新任务提交导致被拒绝。