调用OpenFGA的CompletableFuture方法致ForkJoinPool线程无法复用引发OOM

问题现象 #

应用运行过程中创建了数千个ForkJoinPool线程池及线程，这些线程始终处于WAIT状态无法关闭，最终触发如下OOM错误（程序未崩溃）：

jakarta.servlet.ServletException: Handler dispatch failed: java.lang.OutOfMemoryError: unable to create native thread: possibly out of memory or process/resource limits reached
// ...stacktrace
[95433.690s][warning][os,thread] Failed to start thread "Unknown thread" - pthread_create failed (EAGAIN) for attributes: stacksize: 1024k, guardsize: 0k, detached.
[95433.691s][warning][os,thread] Failed to start the native thread for java.lang.Thread "pool-1917-thread-1"
Caused by: java.lang.OutOfMemoryError: unable to create native thread: possibly out of memory or process/resource limits reached

程序中仅OpenFGA库使用ForkJoinPool，核心调用代码如下：

// 业务调用代码
var response = fgaClient.listRelations(body, options).get();

// OpenFGA库中listRelations方法实现片段
public CompletableFuture<ClientListRelationsResponse> listRelations(ClientListRelationsRequest request, ClientListRelationsOptions listRelationsOptions) throws FgaInvalidParameterException {
  // 实现代码...

  return this.batchCheck(batchCheckRequests, options.asClientBatchCheckOptions())
                .thenCompose(responses -> call(() -> ClientListRelationsResponse.fromBatchCheckResponses(responses)));
}

问题原因 #

这是OpenFGA库的已知bug：在0.7.2版本之前，库内部的异步方法（如listRelations依赖的batchCheck等）每次调用时都会创建新的ForkJoinPool实例，且这些线程池没有被正确配置为可自动回收或手动关闭，导致线程池中的线程长期处于WAIT状态无法被JVM回收。随着调用次数增加，线程池和线程数量持续累积，最终耗尽系统的线程资源，触发“unable to create native thread”类型的OOM错误。

修复方案 #

直接将OpenFGA库升级到0.7.2及以上版本，该版本已修复线程池泄漏的问题，确保ForkJoinPool实例被合理复用或正确回收，避免无限制创建线程池和线程。

内容的提问来源于stack exchange，提问作者Minh Long Vu