I have a number of Runnable tasks governed by an executor service.
These tasks are essentially JMS queue pollers and contain their own connections.
For example:
ExecutorService executor = Executors.newFixedThreadPool(...);
executor.submit(new MyListener());
My listener:
public class MyListener implements Runnable {
#Override
public void run() {
// Create my JMS connection here
}
}
How do I gracefully close the JMS connection in each task and then proceed to shutdown each thread?
I'm having problems shutting down the executor service with shutdown().
I need to force an interrupt with shutdownNow(), however, how can I be sure that my JMS connection has been closed without me explicitly calling .close()?
Is there an interface I'm missing that allows shutdown hooks to be executed when I attempt to stop the task?
Here's my solution to gracefully shutdown threads holding a connection, utilising the suggestion of an atomic boolean:
MyRunnable implements Runnable {
private volatile boolean isActive = true;
public void run() {
while(isActive && !Thread.currentThread().isInterrupted()) {
...
}
}
public void stop() {
isActive = false;
}
}
Main thread:
private void myShutdownHook() {
myRunnableInstance.stop();
// Can safely shutdown executor now...
}
The easiest would be to use an atomic boolean shared by the class that signals shutdown and the runnable. This tells them to stop. Another option is finding the threads yourself and interrupt them. You would need to catch the interrupted exception and close the connection. You can set the thread names when the runnable launch for easy id.
Related
In the book, Java Concurrency in Practice by Brian Goetz et al, the example on page 141 (2006):
7.5: Using interruption for cancellation.
class PrimeProducer extends Thread {
}
...
public void cancel() { interrupt(); }
The confusing thing is that the book states that Threads should implement an Interruption Policy, while Runnable / Callable tasks should implement a Cancellation Policy.
Yet here we are with a cancel() method inside of a Thread object. What's up with that? A few pages before, an example with Runnable is given (7.1) with cancel(). In the case of tasks, I would expect to see a qualified interrupt() like this:
public void cancel() { Thread.currentThread().interrupt(); }
Extra, semi-relevant information
I am using an ExecutorService, so I deal with tasks (not threads--except for a thread factory for the ExecutorService), but I could not find any could examples of a full ExecutorService shutdown (of many threads) in the book.
My methods for starting tasks and stopping them are:
Map<CancellableRunnable, Future<?>> cancellableFutures = new HashMap<>(); // keep track of refs to tasks for stop()
public void init() {
Future<?> future = myExecutorService.submit(myTask);
cancellableFutures.put(myTask, future);
}
public void stop() {
for (Future task : cancellableFutures.values()) {
task.cancel(true); // also a confusing step. Should it be cancel() on Future or cancel() on task (Runnable/Callable)?
}
}
The confusing thing is that the book states that Threads should implement an Interruption Policy
Right,
class MyThread extends Thread {
#Override
public void interrupt() { ... }
}
while Runnable / Callable tasks should implement a Cancellation Policy.
Right,
// FutureTask = Runnable (for run) + Future<Void> (for cancel(boolean))
class MyTask extends FutureTask<Void> {
#Override
public boolean cancel(boolean mayInterruptIfRunning) { ... }
#Override
public void run() { ... }
}
Yet here we are with a cancel() method inside of a Thread object.
Thread is both Thread and Runnable, so both interrupt (to interrupt this thread) and cancel (to cancel this task, the task currently being run by this thread) should be defined.
public class Thread implements Runnable { ... }
The PrimeProducer example is a bit confusing because it assumes the task defined in PrimeProducer will be used outside PrimeProducer.
class PrimeProducer extends Thread {
public void run() {
try {
BigInteger p = BigInteger.ONE;
while (!Thread.currentThread().isInterrupted())
queue.put(p = p.nextProbablePrime());
} catch (InterruptedException consumed) {
/* Allow thread to exit */
}
}
public void cancel() { interrupt(); }
}
It's very reasonable and accurate since we can do
Runnable runnable = new PrimeProducer();
new Thread(runnable).start();
It's rarely the case, though. It's highly likely we would simply go with
new PrimeProducer().start();
which would make the task we define in run context-aware and Thread.currentThread().isInterrupted() and isInterrupted() would mean the same. That's what your confusion over Thread.currentThread().interrupt() and interrupt() comes from.
In the case of tasks, I would expect to see a qualified interrupt() like this:
public void cancel() { Thread.currentThread().interrupt(); }
That interrupts your own thread, not the thread running the task. There's no point in interrupting yourself if you want something else to stop what it's doing: you can simply stop what you're doing instead.
(You might interrupt the current thread, for example, if you have just caught an InterruptedException, and want to preserve the fact that the thread was interrupted. But you don't use this as a mechanism to start the interruption).
To correctly close a thread, you have to ask it to close itself by calling thread.interrupt() and the thread should periodically check thread.isInterrupted() method.
See more details in official documentation.
For your example, you have an ExecutorService myExecutorService. To close all submitted threads (along with thread pool itself), you could call myExecutorService.shutdown(). As a result, the thread pool calls thread.interrupt() for all threads.
To stop required threads only, you do correct calling future.cancel(true). In this case, your thread pool will be alive and will able to submit another task.
I am running simple thread which has run method as follows
public run()
while(!stopFlag){
// print something Line 1
// print something Line 2
// print something Line 3
// print something Line 4
}
If I run this thread through ExecutorService viz
ExecutorService exs = Executors.newFixedThreadPool(5);
exs.execute(new MyThread));
I stop the ExecutorService
exs.shutdown();
But this does not stop the thread as flag is not set to false. In another question related to same topic I was asked to properly handle InterruptedException which is caused when exs.shutdown() is called.
But in this case I am not doing any action that can throw InterruptedException.
What is the standard way to handle such case ?
Further question
Answer given by Sabir says "If your runnable doesn't respond well to interrupts, nothing can be done to stop it other than shutting down the JVM. ".This seems to be my case.
But how to introduce handling of InterruptedException; if I am not calling any method that throws interrupted exception?
If you are willing to shut your thread even if that flag remains true, you should use - ExecutorService.shutdownNow() method instead of ExecutorService.shutdown()
Quoting from Java Docs,
shutdown()
Initiates an orderly shutdown in which previously submitted tasks are
executed, but no new tasks will be accepted. Invocation has no
additional effect if already shut down.
This method does not wait for previously submitted tasks to complete
execution. Use awaitTermination to do that.
shutdownNow()
Attempts to stop all actively executing tasks, halts the processing of
waiting tasks, and returns a list of the tasks that were awaiting
execution.
This method does not wait for actively executing tasks to terminate.
Use awaitTermination to do that.
There are no guarantees beyond best-effort attempts to stop processing
actively executing tasks. For example, typical implementations will
cancel via Thread.interrupt, so any task that fails to respond to
interrupts may never terminate.
For standard way, I will quote from JDK example from ExecutorService interface,
Usage Examples
Here is a sketch of a network service in which threads in a thread pool service incoming requests. It uses the preconfigured Executors.newFixedThreadPool factory method: class NetworkService implements Runnable { private final ServerSocket serverSocket; private final ExecutorService pool;
public NetworkService(int port, int poolSize)
throws IOException {
serverSocket = new ServerSocket(port);
pool = Executors.newFixedThreadPool(poolSize); }
public void run() { // run the service
try {
for (;;) {
pool.execute(new Handler(serverSocket.accept()));
}
} catch (IOException ex) {
pool.shutdown();
} } }
class Handler implements Runnable { private final Socket socket; Handler(Socket socket) { this.socket = socket; } public void run() {
// read and service request on socket } }} The following method shuts down an ExecutorService in two phases, first by calling shutdown to reject incoming tasks, and then calling shutdownNow, if necessary, to cancel any lingering tasks: void shutdownAndAwaitTermination(ExecutorService pool) { pool.shutdown(); // Disable new tasks from being submitted try {
// Wait a while for existing tasks to terminate
if (!pool.awaitTermination(60, TimeUnit.SECONDS)) {
pool.shutdownNow(); // Cancel currently executing tasks
// Wait a while for tasks to respond to being cancelled
if (!pool.awaitTermination(60, TimeUnit.SECONDS))
System.err.println("Pool did not terminate");
} } catch (InterruptedException ie) {
// (Re-)Cancel if current thread also interrupted
pool.shutdownNow();
// Preserve interrupt status
Thread.currentThread().interrupt(); } }}
Notice that there are no guarantees even with shutdownNow() .
EDIT : If I change your while(!stopFlag) to while(!Thread.currentThread().isInterrupted()) then thread with conditional loop get shutdown with shutdownNow() but not with shutdown() so thread gets interrupted with shutdownNow(). I am on JDK8 and Windows 8.1. I do have to put a sleep in main thread so that service can get time to set up the service and launch runnable. Thread gets launched, goes in while then stops when shutdownNow() is called. I don't get that behavior with shutdown() i.e. thread never comes out of while loop. So the approach to make your runnables responsible for interrupts should be there ,either by checking flags or handling exceptions. If your runnable doesn't respond well to interrupts, nothing can be done to stop it other than shutting down the JVM.
One good approach is shown here
well from your question I am assuming that you are trying to shutdown the process gracefully. In order to do so you need to register a shutdownHook to achieve it. Here is a sample code to achieve it.
package com.example;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
public class ThreadManager {
public static void main(String[] args) {
MyThread myThread = new MyThread();
Runtime.getRuntime().addShutdownHook(new Thread(){
MyThread myThread = null;
#Override
public void run(){
System.out.println("Shutting down....");
this.myThread.stopProcess();
}
public Thread setMyThread(MyThread myThread){
this.myThread=myThread;
return this;
}
}.setMyThread(myThread));
ExecutorService exs = Executors.newFixedThreadPool(5);
myThread.setName("User");
exs.execute(myThread);
exs.shutdownNow();
}
}
And in MyThread.java will be look like following:-
package com.example;
public class MyThread extends Thread{
private boolean stopFlag;
#Override
public void run(){
while(!stopFlag){
System.out.println(this.getName());
}
}
public void stopProcess(){
this.stopFlag=true;
}
}
Now if you make a jar file of this code and run the in a Linux server to see how it is working, then follow these additional steps
Step 1> nohup java -jar MyThread.jar &
Press ctrl+c to exist
Now find the pid using following command
Step 2> ps -ef| grep MyThread.jar
Once you got the pid than execute the following command to stop gracefully
Step 3>kill -TERM <Your PID>
When you check the nohub.out file, the output will looks something like following
User
User
.
.
.
User
Shutting down....
User
.
.
Remember if you try to shutdown using kill -9 than you will never see the Shutting down.... message.
#Sabir already discuss the difference between shutdown and shutdownNow. However I will never recommend you to use interrupt call while the threads are running. It might cause memory leak in real time environment.
Upadte 1:-
public static void main(String[] args) {
MyThread myThreads[] = new MyThread[5];
ExecutorService exs = Executors.newFixedThreadPool(5);
for(int i=0;i<5;++i){
MyThread myThread = new MyThread();
myThread.setName("User "+i);
exs.execute(myThread);
myThreads[i] = myThread;
}
Runtime.getRuntime().addShutdownHook(new Thread(){
MyThread myThreads[] = null;
#Override
public void run(){
System.out.println("Shutting down....");
for(MyThread myThread:myThreads){
myThread.stopProcess();
}
}
public Thread setMyThread(MyThread[] myThreads){
this.myThreads=myThreads;
return this;
}
}.setMyThread(myThreads));
exs.shutdownNow();
}
In my code I'm using CompletionService and ExecutorService in order to start a bunch of Thread to execute some task (that can take a lot of time).
So I have a method that creates the ExecutorService and the CompletionService, then starts submitting threads and then take the results.
I would like to add a shutdown hook in order to gracefully shutdown the executor (I know that probably I should handle releasing resources instead of executor shutdown but in my case each thread has its own resources so shutting down them gracefully can be a good soulution I suppose).
For this reason I write the following code
public Class myClass{
...
private CompletionService<ClusterJobs> completion;
final long SHUTDOWN_TIME = TimeUnit.SECONDS.toSeconds(10);
...
public Message executeCommand(Message request){
final ExecutorService executor = Executors.newFixedThreadPool(30);
completion = new ExecutorCompletionService<ClusterJobs>(executor);
....//submit and take results
Runtime.getRuntime().addShutdownHook(new Thread(){
#Override
public void run() {
logger.debug("Shutting down executor");
try {
if (!executor.awaitTermination(SHUTDOWN_TIME, TimeUnit.SECONDS)) {
logger.debug("Executor still not terminate after waiting time...");
List<Runnable> notExecuted= executor.shutdownNow();
logger.debug("List of dropped task has size " + droppedTasks.size());
}
}catch(InterruptedException e){
logger.error("",e);
}
}
});
}
}
Do you think that this is a reasonable solution or it's unsafe to register and unregister shutdown hook using local classes?
Thanks in advance
Regards
From Design of the Shutdown Hooks API:
Simple shutdown hooks can often be written as anonymous inner classes, as in this example:
Runtime.getRuntime().addShutdownHook(new Thread() {
public void run() { database.close(); }
});
This idiom is fine as long as you'll never need to cancel the hook, in which case you'd need to save a reference to the hook when you create it.
I have a main thread that creates several threads using Executors
ExecutorService executor = Executors.newFixedThreadPool(4);
Each thread has long running jobs (some legacy code from another team) which might run for hours.
Now I want to shutdown from the main thread using
executor.shutdownNow()
And I want the threads to be able to stop immediately, how could I do that?
In the thread, say we have such code:
public void run() {
doA();
doB();
doC();
...
...
}
Now my issue is, even if I called shutdownNow, the running thread will run to the end then stop. I'd like to know how to stop and exit.
It's a slightly tricky situation indeed!
Can we make use of a hook that the JDK has provided in the form of ThreadFactory that is consulted when the associated thread pool is creating a thread in which your legacy task will run? If yes, then why not make your legacy code run in a daemon thread? We know that the JVM exits when the last non-daemon thread exits. So, if we make each thread that the thread pool uses to run your legacy tasks a daemon thread, there is a chance that we can make the shutdownNow() call more responsive:
public class LegacyCodeExecutorEx {
public static void main(String[] args) throws InterruptedException {
ExecutorService executor = Executors.newFixedThreadPool(2, new DaemonThreadFactory());
executor.submit(new LegacySimulator());
Thread.sleep(1000);
executor.shutdownNow();
}
static class LegacySimulator implements Runnable {
private final AtomicLong theLong;
LegacySimulator() {
theLong = new AtomicLong(1);
}
#Override
public void run() {
for (long i = 10; i < Long.MAX_VALUE; i++) {
theLong.set(i*i);
}
System.out.println("Done!");
}
}
static class DaemonThreadFactory implements ThreadFactory {
#Override
public Thread newThread(Runnable r) {
Thread t = new Thread(r);
t.setName("Daemon Thread");
t.setDaemon(true);
return t;
}
}
}
If you play with setDaemon(true) line, you will see that this code either responds to the exit of the main thread (which is non-daemon) either immediately or takes its own sweet time to finish the task.
Is making your legacy-code-running threads daemon threads a possibility? If yes, you could give this a try.
You need to include a flag in the Runnable object instantiation that checks between tasks whether you need to stop or not.
public void run() {
if(timeToShutdown) return;
doA();
if(timeToShutdown) return;
doB();
/*etc*/
}
Threads in Java operate at a (relatively) low level. Short of directly shutting down the entire JVM, the only way to manually force the stop of a Thread is using Deprecated behavior from Java 1.0/1.1, which pretty much noone wants you to use.
I want to achieve the following: When my application starts, the main thread will start 1+ worker threads that should run in the background, and periodically do things behind the scenes. These should not block the main thread: once main starts the workers, it continues doing its own thing until:
The main thread finishes (normal application termination) - in the case of a command-line utility this is when the end of the main(String[]) method is reached; in the case of a Swing GUI it could be when the user selects the File >> Exit menu, etc.
The operating system throws a kill command (SIGKILL, etc.)
An unexpected, uncaught exception occurs in the main thread, effectively killing it (this is just an unpolite version of #1 above)
Once started/submitted from the main thread, I want all the worker threads (Runnables) to essentially have their own life cycle, and exist independently of the main thread. But, if the main thread dies at any time, I want to be able to block (if at all possible) the main thread until all the workers are finished shutting down, and then "allow" the main thread to die.
My best attempt so far, although I know I'm missing pieces here and there:
public class MainDriver {
private BaneWorker baneWorker;
private ExecutorService executor = Executors.newCachedThreadPool();
public static void main(String[] args) {
MainDriver driver = new MainDriver();
driver.run();
// We've now reached the end of the main method. All workers should block while they shutdown
// gracefully (if at all possible).
if(executor.awaitTermination(30, TimeUnit.SECONDS))
System.out.println("Shutting down...");
else {
System.out.println("Forcing shut down...");
executor.shutdownNow();
}
}
private void run() {
// Start all worker threads.
baneWorker = new BaneWorker(Thread.currentThread());
// More workers will be used once I get this simple example up and running...
executor.submit(baneWorker);
// Eventually submit the other workers here as well...
// Now start processing. If command-line utility, start doing whatever the utility
// needs to do. If Swing GUI, fire up a parent JFrame and draw the application to the
// screen for the user, etc.
doStuff();
}
private void doStuff() {
// ??? whatever
}
}
public class BaneWorker implements Runnable {
private Timer timer;
private TimerTask baneTask;
private Thread mainThread;
public BaneWorker(Thread mainThread) {
super();
this.mainThread = mainThread;
}
#Override
public void run() {
try {
timer = new Timer();
baneTask = new TimerTask() {
#Override
public void run() {
System.out.println("When the main thread is ashes...");
}
};
// Schedule the baneTask to kick off every minute starting now.
timer.scheduleAtFixedRate(baneTask, new Date(), 60 * 1000);
} catch(InterruptedException interrupt) {
// Should be thrown if main thread dies, terminates, throws an exception, etc.
// Should block main thread from finally terminating until we're done shutting down.
shutdown();
}
}
private void shutdown() {
baneTask.cancel();
System.out.println("...then you have my permission to die.");
try {
mainThread.join();
} catch(InterruptedException interrupt) {
interrupt.printStackTrace;
}
}
}
Am I on-track or way off-base here? What do I need to change to make this work the way I need it to? I'm new to Java concurrency and am trying my best to use the Concurrency API correctly, but stumbling around a bit. Any ideas? Thanks in advance!
The main thread must signal the worker threads to terminate (generally this is achieved just by using a flag) and then it should call join on every thread to wait for their termination. Have a look here: Java: How to use Thread.join
You can use Runtime.addShutdownHook to register an un-started thread that is executed when a JVM is terminated, the system is shutting down etc. This code can do some cleanup itself, or perhaps notify running daemon threads to finish their work. Any such cleanup code must be relatively fast, because on many systems programs have only a limited time to do cleanup before they're forcibly terminated.
Perhaps you could also consider making your background thread daemon threads. Then they will not block the JVM when main finishes and will be still running during the clean-up phase.
Note that you can't intercept SIGKILL - this signal is designed to be unavoidable and immediate. But it should work with SIGTERM, SIGHUP and similar signals.
Update: You can easily create ExecutorServices that run daemon threads. All you need is to create a proper ThreadFactory:
public static class DaemonFactory
implements ThreadFactory
{
#Override
public Thread newThread(Runnable r) {
Thread t = new Thread(r);
t.setDaemon(true);
return t;
}
}
than you create an ExecutorService like
public static void main(String argv[])
throws Exception
{
ExecutorService es
= Executors.newCachedThreadPool(new DaemonFactory());
// ^^^^^^^^^^^^^^^^^^^
es.submit(new Callable<Object>() {
public Object call() throws Exception {
Thread.sleep(100);
System.err.println("Daemon: " +
Thread.currentThread().isDaemon());
return null;
}
});
// Without this, JVM will terminate before the daemon thread prints the
// message, because JVM doesn't wait for daemon threads when
// terminating:
es.awaitTermination(3, TimeUnit.SECONDS);
}
Concerning Thread.join(), you shouldn't try to use it on threads managed by an ExecutorService. It's the responsibility of the executor to manage them. You have no reliable way how to enumerate its threads, the executor can create and destroy threads depending on its configuration etc. The only reliable way is to call shutdown(); and then awaitTermination(...);.
If SIGKILL is a unix "kill -9" there's nothing you can do about it.
For graceful exits, use a try/catch/finally in your main. The catch will catch your exceptions and allow you to do what needs to be done (recover? abort?) The finally will give you the hook to spin down your threads gracefully.
Reviewing your code quickly, I don't see where you're keeping track of your thread instances. You'll need those if you're going to tell them to spin down.
psuedocode:
static Main(...) {
ArrayList threads = new ArrayList();
try {
for (each thread you want to spin up) {
threads.add(a new Thread())
}
}
catch { assuming all are fatal. }
finally {
for(each thread t in threads) {
t.shutdown();
t.join(); /* Be prepared to catch (and probably ignore) an exception on this, if shutdown() happens too fast! */
}
}