MappedByteBufferSource's ASYNC_FLUSH_EXECUTOR never shut down

[ljacomet]

org.terracotta.offheapstore.disk.paging.MappedPageSource.ASYNC_FLUSH_EXECUTOR is an instance of org.terracotta.offheapstore.util.Retryer and owns an executor which is not shut down.

This is my ugly workaround:

Field retryerField = MappedPageSource.class.getDeclaredField("ASYNC_FLUSH_EXECUTOR");
retryerField.setAccessible(true);

Retryer retryer = (Retryer) retryerField.get(null);

Field executorField = Retryer.class.getDeclaredField("executor");
ScheduledThreadPoolExecutor executor = (ScheduledThreadPoolExecutor) executorField.get(retryer);

executor.shutdown();

Copied from ehcache/ehcache3#1444

[chrisdennis]

The likely way to fix this is to not shutdown the executor on "final shutdown", but instead shut it down when it becomes idle, and then get restarted again on the next usage.

[sgup432]

I think this issue is related to https://github.com/ehcache/ehcache3/issues/3204

[sgup432]

I am guessing we should shutdown the executor like below? @chrisdennis

class MappedPagedSource {
  public synchronized void close() throws IOException {
        try {
            this.channel.close();
            ASYNC_FLUSH_EXECUTOR.shutdown(); // shutdown the retryer like this?
        } finally {
            this.raf.close();
        }

    }
}