在 Worker 执行重任务时，`CommonTaskTracker` 的定时线程正在访问 `DbTaskPersistenceService` / H2 数据库。如果此时触发 `HeavyTaskTracker.destroy()`，其中会对 `scheduledPool` 执行 `shutdownNow()`，可能中断正在执行 H2 SQL 的线程，导致 H2 数据库被关闭或文件异常。

org.h2.jdbc.JdbcSQLNonTransientConnectionException: The database has been closed [90098-214]
Caused by: org.h2.mvstore.MVStoreException: Reading from file ... failed
Caused by: java.nio.channels.ClosedChannelException

想确认：这里是否存在因 shutdownNow() 中断正在执行数据库 I/O 的线程，导致 H2 数据库被关闭甚至文件异常的风险？

相关代码位置

HeavyTaskTracker.destroy()

// tech.powerjob.worker.core.tracker.task.heavy.HeavyTaskTracker

@Override
public void destroy() {
    finished.set(true);

    // 0. 开始关闭线程池，不能使用 shutdownNow()，因为 destroy 方法本身就在 scheduledPool 的线程中执行，强行关闭会打断 destroy 的执行。
    scheduledPool.shutdown();

    // ...

    // 2. 删除所有数据库数据
    boolean dbSuccess = taskPersistenceService.deleteAllTasks(instanceId);

    // ...

    // 4. 强制关闭线程池
    if (!scheduledPool.isTerminated()) {
        CommonUtils.executeIgnoreException(() -> scheduledPool.shutdownNow());
    }
}

CommonTaskTracker 状态检查线程

// tech.powerjob.worker.core.tracker.task.heavy.CommonTaskTracker

scheduledPool.scheduleWithFixedDelay(new StatusCheckRunnable(), 3, delay, TimeUnit.SECONDS);

StatusCheckRunnable 中会查询任务状态：

InstanceStatisticsHolder holder = getInstanceStatisticsHolder(instanceId);

最终调用：

taskPersistenceService.getTaskStatusStatistics(instanceId, subInstanceId);

DbTaskPersistenceService

private <T> T execute(SupplierPlus<T> executor, Consumer<Long> slowQueryLogger) throws Exception {
    long s = System.currentTimeMillis();
    try {
        return CommonUtils.executeWithRetry(executor, RETRY_TIMES, RETRY_INTERVAL_MS);
    } finally {
        long cost = System.currentTimeMillis() - s;
        if (cost > SLOW_QUERY_RT_THRESHOLD) {
            slowQueryLogger.accept(cost);
        }
    }
}

TaskDAOImpl.simpleQueryPlus

try (Connection conn = connectionFactory.getConnection(); PreparedStatement ps = conn.prepareStatement(sql)) {
    rs = ps.executeQuery();
    ...
}

异常堆栈

[06-12 05:30:58.812] ERROR [] powerjob-worker-batch ctttp-945530341840388544-0 tech.powerjob.worker.persistence.DbTaskPersistenceService [] [TaskPersistenceService] getTaskStatusStatistics for instance(id=945530341840388544) failed.
org.h2.jdbc.JdbcSQLNonTransientConnectionException: The database has been closed [90098-214]
    at org.h2.message.DbException.getJdbcSQLException(DbException.java:678)
    at org.h2.message.DbException.getJdbcSQLException(DbException.java:477)
    at org.h2.message.DbException.get(DbException.java:212)
    at org.h2.engine.SessionLocal.getTransaction(SessionLocal.java:1596)
    at org.h2.engine.SessionLocal.startStatementWithinTransaction(SessionLocal.java:1617)
    at org.h2.command.Command.executeQuery(Command.java:184)
    at org.h2.jdbc.JdbcPreparedStatement.executeQuery(JdbcPreparedStatement.java:128)
    at com.zaxxer.hikari.pool.ProxyPreparedStatement.executeQuery(ProxyPreparedStatement.java:52)
    at com.zaxxer.hikari.pool.HikariProxyPreparedStatement.executeQuery(HikariProxyPreparedStatement.java)
    at tech.powerjob.worker.persistence.db.TaskDAOImpl.simpleQueryPlus(TaskDAOImpl.java:129)
    at tech.powerjob.worker.persistence.DbTaskPersistenceService.lambda$getTaskStatusStatistics$16(DbTaskPersistenceService.java:287)
    at tech.powerjob.common.utils.CommonUtils.executeWithRetry(CommonUtils.java:47)
    at tech.powerjob.worker.persistence.DbTaskPersistenceService.execute(DbTaskPersistenceService.java:399)
    at tech.powerjob.worker.persistence.DbTaskPersistenceService.getTaskStatusStatistics(DbTaskPersistenceService.java:286)
    at tech.powerjob.worker.persistence.SwapTaskPersistenceService.getTaskStatusStatistics(SwapTaskPersistenceService.java:174)
    at tech.powerjob.worker.core.tracker.task.heavy.HeavyTaskTracker.getInstanceStatisticsHolder(HeavyTaskTracker.java:440)
    at tech.powerjob.worker.core.tracker.task.heavy.CommonTaskTracker$StatusCheckRunnable.innerRun(CommonTaskTracker.java:165)
    at tech.powerjob.worker.core.tracker.task.heavy.CommonTaskTracker$StatusCheckRunnable.run(CommonTaskTracker.java:322)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
    at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:308)
    at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:180)
    at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:294)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
    at java.lang.Thread.run(Thread.java:855)
Caused by: org.h2.mvstore.MVStoreException: Reading from file sun.nio.ch.FileChannelImpl@e0a6f13 failed at 7069598499 (length -1), read 0, remaining 8192 [2.1.214/1]
    at org.h2.mvstore.DataUtils.newMVStoreException(DataUtils.java:1004)
    at org.h2.mvstore.DataUtils.readFully(DataUtils.java:470)
    at org.h2.mvstore.FileStore.readFully(FileStore.java:98)
    at org.h2.mvstore.Chunk.readBufferForPage(Chunk.java:422)
    at org.h2.mvstore.MVStore.readPage(MVStore.java:2569)
    at org.h2.mvstore.MVMap.readPage(MVMap.java:633)
    at org.h2.mvstore.Page$NonLeaf.getChildPage(Page.java:1125)
    at org.h2.mvstore.CursorPos.traverseDown(CursorPos.java:61)
    at org.h2.mvstore.MVMap.operate(MVMap.java:1770)
    at org.h2.mvstore.MVMap.put(MVMap.java:156)
    at org.h2.mvstore.MVStore.acceptChunkOccupancyChanges(MVStore.java:1894)
    at org.h2.mvstore.MVStore.serializeToBuffer(MVStore.java:1675)
    at org.h2.mvstore.MVStore.serializeAndStore(MVStore.java:1598)
    at org.h2.mvstore.MVStore.lambda$storeNow$4(MVStore.java:1518)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
    at java.util.concurrent.FutureTask.run(FutureTask.java:266)
    ... 3 common frames omitted
Caused by: java.nio.channels.ClosedChannelException
    at sun.nio.ch.FileChannelImpl.ensureOpen(FileChannelImpl.java:110)
    at sun.nio.ch.FileChannelImpl.read(FileChannelImpl.java:715)
    at org.h2.mvstore.DataUtils.readFully(DataUtils.java:456)
    ... 17 common frames omitted

初步分析

从堆栈看，异常发生在 ctttp-{instanceId}-0 线程中，该线程属于 CommonTaskTracker 的 scheduledPool。

该线程正在执行：

StatusCheckRunnable
→ getInstanceStatisticsHolder
→ getTaskStatusStatistics
→ TaskDAOImpl.simpleQueryPlus
→ PreparedStatement.executeQuery
→ H2 MVStore FileChannel read

如果此时 HeavyTaskTracker.destroy() 执行到：

scheduledPool.shutdownNow();

则会对 scheduledPool 中正在运行的线程执行 interrupt。

Java NIO 的 FileChannel 属于可中断通道，线程在执行文件 I/O 时被 interrupt，可能导致 channel 被关闭，从而触发：

java.nio.channels.ClosedChannelException

随后 H2 MVStore 抛出：

MVStoreException: Reading from file ... failed

最终表现为：

JdbcSQLNonTransientConnectionException: The database has been closed

疑问

1. HeavyTaskTracker.destroy() 中最后的 scheduledPool.shutdownNow() 是否可能中断正在执行 H2 SQL 的状态检查线程？
2. H2 MVStore 在被 interrupt 后出现 ClosedChannelException，是否可能导致整个 Worker 内置数据库被标记为 closed？
3. 这里是否应该避免在 destroy() 中使用 shutdownNow()，改为更温和的关闭方式，例如：
  - 仅使用 scheduledPool.shutdown()；
  - 或先 shutdown()，再 awaitTermination()；
  - 超时后再根据场景决定是否 shutdownNow()；
  - 或确保数据库访问完全结束后再中断线程池。
4. CommonUtils.executeWithRetry() 中目前会吞掉第一次 executor.get() 的异常，如果随后 Thread.sleep() 被 interrupt，日志中可能只看到 sleep interrupted，是否也需要改进异常保留逻辑？

期望行为

当任务结束或 TaskTracker.destroy() 执行时，不应因为强制中断正在执行数据库 I/O 的线程，导致 H2 数据库被关闭或文件异常。

实际行为

StatusCheckRunnable 正在查询 H2 时，疑似被 shutdownNow() 中断，出现：

The database has been closed [90098-214]
Caused by: MVStoreException
Caused by: ClosedChannelException

环境信息

- PowerJob 版本：5.1.1
- H2 版本：2.1.214
- JDK 版本：8
- Worker 存储策略：DISK
- 任务类型：MAP_REDUCE
- 是否使用 SWAP：是

可能的修复方向

建议考虑：

1. HeavyTaskTracker.destroy() 不再直接 shutdownNow()；
2. 对 scheduledPool 使用优雅关闭流程；
3. DbTaskPersistenceService / CommonUtils.executeWithRetry() 对 InterruptedException 做特殊处理，恢复中断标记并保留原始 SQL 异常；
4. 在 Worker 销毁时，保证数据库相关线程池先优雅停止，再释放底层 H2/Hikari 资源