[HDFS-12152] [SPS]: Re-arrange StoragePolicySatisfyWorker stopping sequence to improve thread cleanup time

• Type: Sub-task
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: HDFS-10285, 3.2.0
• Component/s: datanode, namenode
• Target Version/s: HDFS-10285
• Hadoop Flags: Reviewed
• Assignee: Rakesh Radhakrishnan
• Reporter: Rakesh Radhakrishnan
• Created: 17/Jul/17 14:34
• Updated: 13/Aug/18 05:40
• Resolved: 19/Jul/17 07:56

Description

This jira to improve the StoragePolicySatisfyWorker#stop sequence of steps to improve the thread interruption and graceful shutdown quickly.

I have observed that TestDataNodeUUID#testUUIDRegeneration test case is getting timed out frequently. When analyzing, it looks like the below function is always taking 3 secs waiting period. Probably, we could improve the thread interruption sequence so that the thread should finish #run method quickly.

StoragePolicySatisfyWorker.java

  void waitToFinishWorkerThread() {
    try {
      movementTrackerThread.join(3000);
    } catch (InterruptedException ignore) {
      // ignore
    }
  }