Commons IO
  1. Commons IO
  2. IO-259

FileAlterationMonitor.stop(boolean allowIntervalToFinish)

    Details

    • Type: Improvement Improvement
    • Status: Closed
    • Priority: Major Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 2.1
    • Component/s: None
    • Labels:
      None

      Description

      I'm a long-time user of commons-io, but I just started using FileAlterationMonitor the other day. I have a bean in a Spring application context that constructs a FileAlterationMonitor, and it sets the ThreadFactory to an instance that creates daemon threads. At context shutdown, my bean calls .stop() on the monitor.

      The problem is that .stop() currently honors the Thread.sleep(interval), in that it does a nice friendly thread.join(). If you set your interval high enough, your "graceful app shutdown" is going to sit there waiting a while.

      Compounding things is that I have seven FileAlterationMonitors in my app, all of which run with a 10-second sleep time. So at graceful shutdown time, I'm facing a delay of up to 70 seconds...lame!

      So I stopped calling .stop() and since the ThreadFactory created daemon threads, shutdown is quick. But...

      I'm running inside tomcat, and when it shuts down it looks for leaks. And of course it finds a handful of my threads and complains, such as:

      SEVERE: The web application [/my-webapp] appears to have started a thread named [FileUpdateMonitor-/path/to/my/file.ext] but has failed to stop it. This is very likely to create a memory leak.

      So what I'm suggesting is an alternate version of FileAlterationMonitor.stop() that takes "boolean allowIntervalToFinish". The default behavior won't change (for backward compatibility). But if you explicitly call .stop(false), then it will interrupt the sleeping thread immediately. That thread wakes up, sees that running=false, and finishes up immediately.

      Patch will be attached in a sec...

      1. IO-259.patch
        2 kB
        Dan Checkoway

        Activity

        Hide
        Gary Gregory added a comment -

        Closing, we released version 2.1.

        Show
        Gary Gregory added a comment - Closing, we released version 2.1.
        Hide
        Dan Checkoway added a comment -

        @Niall, thanks! Much appreciated. Looking forward to 2.1.

        Show
        Dan Checkoway added a comment - @Niall, thanks! Much appreciated. Looking forward to 2.1.
        Hide
        Niall Pemberton added a comment -

        Good suggestions - I added a stop(long) method and if you specify a negative value it stops immediately.

        http://svn.apache.org/viewvc?view=revision&revision=1080843

        Show
        Niall Pemberton added a comment - Good suggestions - I added a stop(long) method and if you specify a negative value it stops immediately. http://svn.apache.org/viewvc?view=revision&revision=1080843
        Hide
        Dan Checkoway added a comment -

        Another variation would be to have FileAlterationMonitor.stop(long millis), and millis would be passed to thread.join(). That way the caller could specify the max allowable wait time for the join. NOTE that 0 means wait forever according to the JDK, so you'd have to pass 1 (or would -1 work?) if you wanted it to wait as little time as possible.

        Show
        Dan Checkoway added a comment - Another variation would be to have FileAlterationMonitor.stop(long millis), and millis would be passed to thread.join(). That way the caller could specify the max allowable wait time for the join. NOTE that 0 means wait forever according to the JDK, so you'd have to pass 1 (or would -1 work?) if you wanted it to wait as little time as possible.
        Hide
        Dan Checkoway added a comment -

        Patch attached, adding the second version of FileAlterationMonitor.stop(boolean). Default behavior is for stop() to call stop(true). But users can override that by calling stop(false) explicitly.

        Please add this to trunk and the next release. Thanks!!

        FWIW, I already have an ICLA on file for my contributions to other projects.

        Show
        Dan Checkoway added a comment - Patch attached, adding the second version of FileAlterationMonitor.stop(boolean). Default behavior is for stop() to call stop(true). But users can override that by calling stop(false) explicitly. Please add this to trunk and the next release. Thanks!! FWIW, I already have an ICLA on file for my contributions to other projects.

          People

          • Assignee:
            Niall Pemberton
            Reporter:
            Dan Checkoway
          • Votes:
            0 Vote for this issue
            Watchers:
            0 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Development