Deadlock occurs during shutdown

[funky-eyes]

Describe the bug

    public void destroy() {
        Optional.ofNullable(raftGroupService).ifPresent(r -> {
            r.shutdown();
            try {
                r.join();
            } catch (InterruptedException e) {
                logger.warn("Interrupted when RaftServer destroying", e);
            }
        });
    }

可以看到jraft的groupshutdown是启动一个新的线程去shutdown，此时NodeImpl的shutdown后也会去执行join方法

"JRaft-Group-Default-Executor-3" #182 [221814] daemon prio=5 os_prio=0 cpu=62441.04ms elapsed=3535428.38s tid=0x00007f4c0411f200 nid=221814 waiting on condition  [0x00007f4bde594000]
   java.lang.Thread.State: WAITING (parking)
	at jdk.internal.misc.Unsafe.park(java.base@21.0.4/Native Method)
	- parking to wait for  <0x00000007017fb0f0> (a java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject)
	at java.util.concurrent.locks.LockSupport.park(java.base@21.0.4/LockSupport.java:371)
	at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionNode.block(java.base@21.0.4/AbstractQueuedSynchronizer.java:519)
	at java.util.concurrent.ForkJoinPool.unmanagedBlock(java.base@21.0.4/ForkJoinPool.java:3780)
	at java.util.concurrent.ForkJoinPool.managedBlock(java.base@21.0.4/ForkJoinPool.java:3725)
	at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(java.base@21.0.4/AbstractQueuedSynchronizer.java:1707)
	at com.alipay.sofa.jraft.util.CountDownEvent.await(CountDownEvent.java:69)
	at com.alipay.sofa.jraft.storage.snapshot.SnapshotExecutorImpl.join(SnapshotExecutorImpl.java:748)
	at com.alipay.sofa.jraft.core.NodeImpl.join(NodeImpl.java:2891)
	- locked <0x00000007017e6d30> (a com.alipay.sofa.jraft.core.NodeImpl)
	at com.alipay.sofa.jraft.core.NodeImpl.lambda$shutdown$7(NodeImpl.java:2837)
	at com.alipay.sofa.jraft.core.NodeImpl$$Lambda/0x00007f4c57804f80.run(Unknown Source)
	at java.util.concurrent.Executors$RunnableAdapter.call(java.base@21.0.4/Executors.java:572)
	at java.util.concurrent.FutureTask.run(java.base@21.0.4/FutureTask.java:317)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(java.base@21.0.4/ThreadPoolExecutor.java:1144)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(java.base@21.0.4/ThreadPoolExecutor.java:642)
	at java.lang.Thread.runWith(java.base@21.0.4/Thread.java:1596)
	at java.lang.Thread.run(java.base@21.0.4/Thread.java:1583)

然后业务线程等待这个结果时，join方法是携带了synchronized，导致shutdownhook线程因为拿不到nodeimpl的锁导致一直hang住，应用无法下线

"SpringApplicationShutdownHook" #34 [442820] prio=5 os_prio=0 cpu=320.57ms elapsed=421.49s tid=0x00007f4bfc1ad980 nid=442820 waiting for monitor entry  [0x00007f4bc0946000]
   java.lang.Thread.State: BLOCKED (on object monitor)
	at com.alipay.sofa.jraft.core.NodeImpl.join(NodeImpl.java:2883)
	- waiting to lock <0x00000007017e6d30> (a com.alipay.sofa.jraft.core.NodeImpl)
	at com.alipay.sofa.jraft.RaftGroupService.join(RaftGroupService.java:148)
	- locked <0x0000000701d33790> (a com.alipay.sofa.jraft.RaftGroupService)
	at org.apache.seata.server.cluster.raft.RaftServer.lambda$destroy$0(RaftServer.java:130)
	at org.apache.seata.server.cluster.raft.RaftServer$$Lambda/0x00007f4c577bf0f8.accept(Unknown Source)
	at java.util.Optional.ifPresent(java.base@21.0.4/Optional.java:178)
	at org.apache.seata.server.cluster.raft.RaftServer.destroy(RaftServer.java:127)
	at org.apache.seata.server.cluster.raft.RaftServer.close(RaftServer.java:122)
	at org.apache.seata.server.cluster.raft.RaftServerManager.lambda$destroy$1(RaftServerManager.java:166)
	at org.apache.seata.server.cluster.raft.RaftServerManager$$Lambda/0x00007f4c577beed0.accept(Unknown Source)
	at java.util.HashMap.forEach(java.base@21.0.4/HashMap.java:1429)
	at org.apache.seata.server.cluster.raft.RaftServerManager.destroy(RaftServerManager.java:165)
	at org.apache.seata.server.session.SessionHolder.destroy(SessionHolder.java:415)
	at org.apache.seata.server.coordinator.DefaultCoordinator.destroy(DefaultCoordinator.java:684)
	at org.apache.seata.server.ServerRunner.destroy(ServerRunner.java:90)
	at org.springframework.beans.factory.support.DisposableBeanAdapter.destroy(DisposableBeanAdapter.java:213)
	at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroyBean(DefaultSingletonBeanRegistry.java:587)
	at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingleton(DefaultSingletonBeanRegistry.java:559)
	at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingleton(DefaultListableBeanFactory.java:1163)
	at org.springframework.beans.factory.support.DefaultSingletonBeanRegistry.destroySingletons(DefaultSingletonBeanRegistry.java:520)
	at org.springframework.beans.factory.support.DefaultListableBeanFactory.destroySingletons(DefaultListableBeanFactory.java:1156)
	at org.springframework.context.support.AbstractApplicationContext.destroyBeans(AbstractApplicationContext.java:1123)
	at org.springframework.context.support.AbstractApplicationContext.doClose(AbstractApplicationContext.java:1089)
	at org.springframework.boot.web.servlet.context.ServletWebServerApplicationContext.doClose(ServletWebServerApplicationContext.java:174)
	at org.springframework.context.support.AbstractApplicationContext.close(AbstractApplicationContext.java:1035)
	- locked <0x0000000700cb6728> (a java.lang.Object)
	at org.springframework.boot.SpringApplicationShutdownHook.closeAndWait(SpringApplicationShutdownHook.java:145)
	at org.springframework.boot.SpringApplicationShutdownHook$$Lambda/0x00007f4c577b5538.accept(Unknown Source)
	at java.lang.Iterable.forEach(java.base@21.0.4/Iterable.java:75)
	at org.springframework.boot.SpringApplicationShutdownHook.run(SpringApplicationShutdownHook.java:114)
	at java.lang.Thread.runWith(java.base@21.0.4/Thread.java:1596)
	at java.lang.Thread.run(java.base@21.0.4/Thread.java:1583)

Expected behavior

Actual behavior

Steps to reproduce

Minimal yet complete reproducer code (or GitHub URL to code)

Environment

• SOFAJRaft version:
• JVM version (e.g. java -version):
• OS version (e.g. uname -a):
• Maven version:
• IDE version:

[funky-eyes]

    @Override
    public synchronized void join() throws InterruptedException {
        if (this.shutdownLatch != null) {
            if (this.readOnlyService != null) {
                this.readOnlyService.join();
            }
            if (this.logManager != null) {
                this.logManager.join();
            }
            if (this.snapshotExecutor != null) {
                this.snapshotExecutor.join();
            }
            if (this.wakingCandidate != null) {
                Replicator.join(this.wakingCandidate);
            }
            this.shutdownLatch.await();
            this.applyDisruptor.shutdown();
            this.applyQueue = null;
            this.applyDisruptor = null;
            this.shutdownLatch = null;
        }
        if (this.fsmCaller != null) {
            this.fsmCaller.join();
        }
    }

snapshot#join 需要等待jobs的触发，而jobs的触发需要做完一次snapshot，而我应用已经下线了，不可能再去做snapshot，所以snapshot就不可能触发，导致直接hang死

[funky-eyes]

NodeImpl的join会去等待snapshot的快照，而shutdown的时候，snapshot不可能去做了，业务线程全部关闭了，我snapshot 10分钟一次，等于起码得等10分钟后才能下线，这太奇怪了，由于raftgroupservice的shutdown行为是异步的，所以我才用join去等待下线，而不是直接暴力下线，导致主线程这里的join在等异步线程的join释放锁，而异步线程的join释放锁需要等snapshot做完

[funky-eyes]

如果这确定是一个bug的话，我认为有2个解决方法
1.在shutdown时，立刻触发做一次snapshot
2.类似logManager增加一个shutdown专用的CountDownLatch，nodeimpl触发shutdown时触发SnapshotExecutorImpl中的CountDownLatch，来使join可通过，而不是等待