I'm comfortable with the idea of orderly shutdown on threads scheduled with an ExectuorService
; that is to say, calling shutdown
or shutdownNow
will cause threads created on the pool to exit gracefully. If they respond to interrupt
you can be sure finally etc will be called and you'll get a clean, predictable exit (where you can cleanup any resources etc).
However, if you've set your thread to be a daemon (via the executor's ThreadFactory
) as below.
ExecutorService pool = Executors.newSingleThreadExecutor(new ThreadFactory() {
@Override
public Thread newThread(Runnable runnable) {
Thread thread = Executors.defaultThreadFactory().newThread(runnable);
thread.setDaemon(true);
return thread;
}
});
after the main thread terminates, the VM will abruptly terminate any daemon threads. In the example above, a (daemon) thread scheduled and then abruptly terminated will bypass any finally blocks and any interruptable methods won't throw InterruptedException
.
So, I tend to think that marking the threads used in a ThreadPoolExecutor
's pool as daemon is bad practice... my question is really about helping me vocalise why.
Why is it bad practice (or not if you disagree) to use daemon threads in a ExecutorService
's thread pool? In particular I'm interested in describing the life-cycle of the VM shutdown with graceful shutdown (threads that have an interruption policy and play nicely) vs daemon threads.
Expanding that last point, finalize
on ThreadPoolExecutor
will call shutdown
on itself, but when it's using daemon threads, they could have terminated already if finalize
was called by the VM. What's the behavior of the thread pool then? Can it be tricked into remaining alive (and so not exiting the VM) if underlying threads terminated abruptly?
Part of the reason I'm asking is because i've seen it used to bypass the need to shutdown the actual ExectorService. Can you think of scenarios where bypassing its shutdown life-cycle can have ill affect? So far, the only reason I can come up with for using daemons is to take a short cut and I want to appreciate any unexpected side affects it could cause.
is it bad practice to use daemon threads in a ExecutorService's thread pool?
If the tasks sent to that particular ExecutorService
are ok to be terminated abruptly, then why not, that's what daemon threads do. But generally, there are not many tasks that are ok to be terminated with no shutdown ceremonies at all, so you must know what you're doing if you opt to daemon threads.
finalize()
is called when an object is about to be garbage collected. There are no guarantees on when, if ever, any particular object will be GCd, and ThreadPoolExecutor
is no exception, so its finalize()
may or may not be called. The behavior depends on the particular JRE implementation, and even with the same implementation, may vary from time to time.