private static void showThreads(PrintStream pw, ThreadGroup g, Thread current) {
int nthreads = g.activeCount();
pw.println("\nThread Group = " + g.getName() + " activeCount= " + nthreads);
Thread[] tarray = new Thread[nthreads];
int n = g.enumerate(tarray, false);
for (int i = 0; i < n; i++) {
Thread thread = tarray[i];
ClassLoader loader = thread.getContextClassLoader();
String loaderName = (loader == null) ? "Default" : loader.getClass().getName();
Thread.State state = thread.getState();
long id = thread.getId();
pw.print(" " + id + " " + thread.getName() + " " + state + " " + loaderName);
if (thread == current) pw.println(" **** CURRENT ***");
else pw.println();
}
int ngroups = g.activeGroupCount();
ThreadGroup[] garray = new ThreadGroup[ngroups];
int ng = g.enumerate(garray, false);
for (int i = 0; i < ng; i++) {
ThreadGroup nested = garray[i];
showThreads(pw, nested, current);
}
}
/// The watcher thread - from the Runnable interface.
// This has to be pretty anal to avoid monitor lockup, lost
// threads, etc.
public synchronized void run() {
Thread me = Thread.currentThread();
me.setPriority(Thread.MAX_PRIORITY);
ThreadMXBean threadMXBean = ManagementFactory.getThreadMXBean();
startTimeInNs = threadMXBean.getCurrentThreadCpuTime();
if (enabled.get()) {
do {
loop.set(false);
try {
wait(millis);
} catch (InterruptedException e) {
}
} while (enabled.get() && loop.get());
}
if (enabled.get() && targetThread.isAlive()) {
isDoneRunning.set(true);
printThread();
if (kill) {
logger.warn(
"Trying to kill thread with id:"
+ targetThread.getId()
+ " but did not as it can cause deadlocks etc.");
// targetThread.interrupt();
// targetThread.stop(); //Never kill thread - it can cause other problems
}
done();
isDoneRunning.set(false);
}
}
public Object down(Event evt) {
switch (evt.getType()) {
case ExecutorEvent.TASK_SUBMIT:
Runnable runnable = evt.getArg();
// We are limited to a number of concurrent request id's
// equal to 2^63-1. This is quite large and if it
// overflows it will still be positive
long requestId = Math.abs(counter.getAndIncrement());
if (requestId == Long.MIN_VALUE) {
// TODO: need to fix this it isn't safe for concurrent modifications
counter.set(0);
requestId = Math.abs(counter.getAndIncrement());
}
// Need to make sure to put the requestId in our map before
// adding the runnable to awaiting consumer in case if
// coordinator sent a consumer found and their original task
// is no longer around
// see https://issues.jboss.org/browse/JGRP-1744
_requestId.put(runnable, requestId);
_awaitingConsumer.add(runnable);
sendToCoordinator(RUN_REQUEST, requestId, local_addr);
break;
case ExecutorEvent.CONSUMER_READY:
Thread currentThread = Thread.currentThread();
long threadId = currentThread.getId();
_consumerId.put(threadId, PRESENT);
try {
for (; ; ) {
CyclicBarrier barrier = new CyclicBarrier(2);
_taskBarriers.put(threadId, barrier);
// We only send to the coordinator that we are ready after
// making the barrier, wait for request to come and let
// us free
sendToCoordinator(Type.CONSUMER_READY, threadId, local_addr);
try {
barrier.await();
break;
} catch (BrokenBarrierException e) {
if (log.isDebugEnabled())
log.debug(
"Producer timed out before we picked up"
+ " the task, have to tell coordinator"
+ " we are still good.");
}
}
// This should always be non nullable since the latch
// was freed
runnable = _tasks.remove(threadId);
_runnableThreads.put(runnable, currentThread);
return runnable;
} catch (InterruptedException e) {
if (log.isDebugEnabled()) log.debug("Consumer " + threadId + " stopped via interrupt");
sendToCoordinator(Type.CONSUMER_UNREADY, threadId, local_addr);
Thread.currentThread().interrupt();
} finally {
// Make sure the barriers are cleaned up as well
_taskBarriers.remove(threadId);
_consumerId.remove(threadId);
}
break;
case ExecutorEvent.TASK_COMPLETE:
Object arg = evt.getArg();
Throwable throwable = null;
if (arg instanceof Object[]) {
Object[] array = (Object[]) arg;
runnable = (Runnable) array[0];
throwable = (Throwable) array[1];
} else {
runnable = (Runnable) arg;
}
Owner owner = _running.remove(runnable);
// This won't remove anything if owner doesn't come back
_runnableThreads.remove(runnable);
Object value = null;
boolean exception = false;
if (throwable != null) {
// InterruptedException is special telling us that
// we interrupted the thread while waiting but still got
// a task therefore we have to reject it.
if (throwable instanceof InterruptedException) {
if (log.isDebugEnabled())
log.debug("Run rejected due to interrupted exception returned");
sendRequest(owner.address, Type.RUN_REJECTED, owner.requestId, null);
break;
}
value = throwable;
exception = true;
} else if (runnable instanceof RunnableFuture<?>) {
RunnableFuture<?> future = (RunnableFuture<?>) runnable;
boolean interrupted = false;
boolean gotValue = false;
// We have the value, before we interrupt at least get it!
while (!gotValue) {
try {
value = future.get();
gotValue = true;
} catch (InterruptedException e) {
interrupted = true;
} catch (ExecutionException e) {
value = e.getCause();
exception = true;
gotValue = true;
}
}
if (interrupted) {
Thread.currentThread().interrupt();
}
}
if (owner != null) {
final Type type;
final Object valueToSend;
if (value == null) {
type = Type.RESULT_SUCCESS;
valueToSend = value;
}
// Both serializable values and exceptions would go in here
else if (value instanceof Serializable
|| value instanceof Externalizable
|| value instanceof Streamable) {
type = exception ? Type.RESULT_EXCEPTION : Type.RESULT_SUCCESS;
valueToSend = value;
}
// This would happen if the value wasn't serializable,
// so we have to send back to the client that the class
// wasn't serializable
else {
type = Type.RESULT_EXCEPTION;
valueToSend = new NotSerializableException(value.getClass().getName());
}
if (local_addr.equals(owner.getAddress())) {
if (log.isTraceEnabled())
log.trace(
"[redirect] <--> ["
+ local_addr
+ "] "
+ type.name()
+ " ["
+ value
+ (owner.requestId != -1 ? " request id: " + owner.requestId : "")
+ "]");
if (type == Type.RESULT_SUCCESS) {
handleValueResponse(local_addr, owner.requestId, valueToSend);
} else if (type == Type.RESULT_EXCEPTION) {
handleExceptionResponse(local_addr, owner.requestId, (Throwable) valueToSend);
}
} else {
sendRequest(owner.getAddress(), type, owner.requestId, valueToSend);
}
} else {
if (log.isTraceEnabled()) {
log.trace("Could not return result - most likely because it was interrupted");
}
}
break;
case ExecutorEvent.TASK_CANCEL:
Object[] array = evt.getArg();
runnable = (Runnable) array[0];
if (_awaitingConsumer.remove(runnable)) {
_requestId.remove(runnable);
ExecutorNotification notification = notifiers.remove(runnable);
if (notification != null) {
notification.interrupted(runnable);
}
if (log.isTraceEnabled())
log.trace("Cancelled task " + runnable + " before it was picked up");
return Boolean.TRUE;
}
// This is guaranteed to not be null so don't take cost of auto unboxing
else if (array[1] == Boolean.TRUE) {
owner = removeKeyForValue(_awaitingReturn, runnable);
if (owner != null) {
Long requestIdValue = _requestId.remove(runnable);
// We only cancel if the requestId is still available
// this means the result hasn't been returned yet and
// we still have a chance to interrupt
if (requestIdValue != null) {
if (requestIdValue != owner.getRequestId()) {
log.warn("Cancelling requestId didn't match waiting");
}
sendRequest(owner.getAddress(), Type.INTERRUPT_RUN, owner.getRequestId(), null);
}
} else {
if (log.isTraceEnabled()) log.warn("Couldn't interrupt server task: " + runnable);
}
ExecutorNotification notification = notifiers.remove(runnable);
if (notification != null) {
notification.interrupted(runnable);
}
return Boolean.TRUE;
} else {
return Boolean.FALSE;
}
case ExecutorEvent.ALL_TASK_CANCEL:
array = evt.getArg();
// This is a RunnableFuture<?> so this cast is okay
@SuppressWarnings("unchecked")
Set<Runnable> runnables = (Set<Runnable>) array[0];
Boolean booleanValue = (Boolean) array[1];
List<Runnable> notRan = new ArrayList<>();
for (Runnable cancelRunnable : runnables) {
// Removed from the consumer
if (!_awaitingConsumer.remove(cancelRunnable) && booleanValue == Boolean.TRUE) {
synchronized (_awaitingReturn) {
owner = removeKeyForValue(_awaitingReturn, cancelRunnable);
if (owner != null) {
Long requestIdValue = _requestId.remove(cancelRunnable);
if (requestIdValue != owner.getRequestId()) {
log.warn("Cancelling requestId didn't match waiting");
}
sendRequest(owner.getAddress(), Type.INTERRUPT_RUN, owner.getRequestId(), null);
}
ExecutorNotification notification = notifiers.remove(cancelRunnable);
if (notification != null) {
log.trace("Notifying listener");
notification.interrupted(cancelRunnable);
}
}
} else {
_requestId.remove(cancelRunnable);
notRan.add(cancelRunnable);
}
}
return notRan;
case Event.SET_LOCAL_ADDRESS:
local_addr = evt.getArg();
break;
case Event.VIEW_CHANGE:
handleView(evt.getArg());
break;
}
return down_prot.down(evt);
}
public boolean runningOnClientThread() {
return clientThread != null && Thread.currentThread().getId() == clientThread.getId();
}
private void printThread() {
long now = System.currentTimeMillis();
long diffLastThreadDump = now - lastThreadDump;
logger.info("diffLastThreadDump:" + diffLastThreadDump);
if (diffLastThreadDump > 60000) {
logger.info("had not sent all threads for a while.. will do so now");
lastThreadDump = now;
try {
ThreadMXBean t = ManagementFactory.getThreadMXBean();
long threads[] = t.getAllThreadIds();
ThreadInfo[] tinfo = t.getThreadInfo(threads, 40);
StringBuilder sb = new StringBuilder("All Threads");
for (int i = 0; i < tinfo.length; i++) {
ThreadInfo e = tinfo[i];
try {
StackTraceElement[] el = e.getStackTrace();
sb.append(
"\n\n"
+ e.getThreadName()
+ "\n"
+ " "
+ " Thread id = "
+ e.getThreadId()
+ " "
+ e.getThreadState());
if (e.getThreadState().equals(State.BLOCKED)) {
sb.append(
"\n\nBlocked info: "
+ e.getBlockedCount()
+ ":"
+ e.getBlockedTime()
+ ":"
+ e.getLockName()
+ ":"
+ e.getLockOwnerId()
+ ":"
+ e.getLockOwnerName()
+ "\n"
+ " "
+ " Thread id = "
+ e.getThreadId()
+ " "
+ e.getThreadState());
ThreadInfo eBlockedThread = t.getThreadInfo(e.getLockOwnerId(), 40);
StackTraceElement[] elBlockedThread = eBlockedThread.getStackTrace();
sb.append(
"\n\n "
+ e.getThreadName()
+ "\n"
+ " "
+ " Thread id = "
+ eBlockedThread.getThreadId()
+ " "
+ eBlockedThread.getThreadState());
if (elBlockedThread == null || elBlockedThread.length == 0) {
sb.append(" no stack trace available");
} else {
for (int n = 0; n < elBlockedThread.length; n++) {
if (n != 0) sb.append("\n");
StackTraceElement frame = elBlockedThread[n];
if (frame == null) {
sb.append(" null stack frame");
continue;
}
sb.append(" ");
sb.append(frame.toString());
}
}
}
if (el == null || el.length == 0) {
sb.append(" no stack trace available");
continue;
}
for (int n = 0; n < el.length; n++) {
if (n != 0) sb.append("\n");
StackTraceElement frame = el[n];
if (frame == null) {
sb.append(" null stack frame");
continue;
}
sb.append(" ");
sb.append(frame.toString());
}
} catch (Exception e2) {
}
}
String warningEmailReceiver = CmsPropertyHandler.getWarningEmailReceiver();
if (warningEmailReceiver != null
&& !warningEmailReceiver.equals("")
&& warningEmailReceiver.indexOf("@warningEmailReceiver@") == -1) {
try {
logger.info("Mailing..");
MailServiceFactory.getService()
.sendEmail(
CmsPropertyHandler.getMailContentType(),
warningEmailReceiver,
warningEmailReceiver,
null,
null,
null,
null,
message,
sb.toString().replaceAll("\n", "<br/>"),
"utf-8");
} catch (Exception e) {
logger.error("Could not send mail:" + e.getMessage(), e);
}
}
} catch (Throwable e) {
logger.error("Error generating message:" + e.getMessage(), e);
}
}
// Only sends if the last stack was sent more than 3 seconds ago.
if ((now - lastSentTimer) > 10000) {
lastSentTimer = System.currentTimeMillis();
StackTraceElement[] el = targetThread.getStackTrace();
ThreadMXBean threadMXBean = ManagementFactory.getThreadMXBean();
long stopTimeInNs = threadMXBean.getThreadUserTime(targetThread.getId());
long diff = (stopTimeInNs - startTimeInNs) / 1000000000;
StringBuffer stackString = new StringBuffer("\n\n" + message + "\n\n");
stackString.append("ServerName: " + getServerName() + "\n");
stackString.append("Maximum memory (MB): " + (maxMemory / 1024 / 1024) + "\n");
stackString.append("Used memory (MB): " + ((totalMemory - freeMemory) / 1024 / 1024) + "\n");
stackString.append("Free memory (MB): " + (freeMemory / 1024 / 1024) + "\n");
stackString.append("Total memory (MB): " + (totalMemory / 1024 / 1024) + "\n");
stackString.append("Number of current requests: " + numberOfCurrentRequests + "\n");
stackString.append("Number of active requests: " + numberOfActiveRequests + "\n");
stackString.append("Number of long requests: " + longThreadMonitorsSize + "\n");
stackString.append("Current thread time: " + diff + " seconds\n");
stackString.append(
"Average time: " + RequestAnalyser.getRequestAnalyser().getAverageElapsedTime() + "\n");
stackString.append(
"Longest time: " + RequestAnalyser.getRequestAnalyser().getMaxElapsedTime() + "\n");
stackString.append("Original url: " + getOriginalFullURL() + "\n");
stackString.append("UserInfo: " + getUserInfo() + "\n");
stackString.append("--------------------------------------------\n\n");
stackString.append("Thread with id [" + targetThread.getId() + "] at report time:\n");
if (el != null && el.length != 0) {
for (int j = 0; j < el.length; j++) {
StackTraceElement frame = el[j];
if (frame == null) stackString.append(" null stack frame" + "\n");
else stackString.append(" ").append(frame.toString()).append("\n");
// if((stackString.indexOf("infoglue") > -1 && j > 20) || j > 35)
// break;
}
}
if (targetThread.getState().equals(State.BLOCKED)) {
ThreadMXBean t = ManagementFactory.getThreadMXBean();
ThreadInfo e = t.getThreadInfo(targetThread.getId(), 40);
stackString.append(
"\n\nBlocked info: "
+ e.getBlockedCount()
+ ":"
+ e.getBlockedTime()
+ ":"
+ e.getLockName()
+ ":"
+ e.getLockOwnerId()
+ ":"
+ e.getLockOwnerName()
+ "\n"
+ " "
+ " Thread id = "
+ e.getThreadId()
+ " "
+ e.getThreadState());
ThreadInfo eBlockedThread = t.getThreadInfo(e.getLockOwnerId(), 40);
StackTraceElement[] elBlockedThread = eBlockedThread.getStackTrace();
stackString.append(
"\n\nBlocked thread: "
+ e.getThreadName()
+ "\n"
+ " "
+ " Thread id = "
+ eBlockedThread.getThreadId()
+ " "
+ eBlockedThread.getThreadState());
if (elBlockedThread == null || elBlockedThread.length == 0) {
stackString.append(" no stack trace available");
} else {
for (int n = 0; n < elBlockedThread.length; n++) {
if (n != 0) stackString.append("\n");
StackTraceElement frame = elBlockedThread[n];
if (frame == null) {
stackString.append(" null stack frame");
continue;
}
stackString.append(" ");
stackString.append(frame.toString());
}
}
}
stackString.append(
"\n\n**********************************\nConcurrent long threads (Only an excerpt of all)\n**********************************");
ThreadMXBean t = ManagementFactory.getThreadMXBean();
List threadMonitors = RequestAnalyser.getLongThreadMonitors();
Iterator threadMonitorsIterator = threadMonitors.iterator();
int threadCount = 0;
while (threadMonitorsIterator.hasNext() && threadCount < 5) {
SimpleThreadMonitor tm = (SimpleThreadMonitor) threadMonitorsIterator.next();
if (targetThread.getId() == tm.getThreadId()) continue;
long threads[] = {tm.getThreadId()};
ThreadInfo[] tinfo = t.getThreadInfo(threads, 40);
stackString
.append("\n\n---------------------------------\nConcurrent long thread [")
.append(tm.getThreadId())
.append("]:\n");
stackString
.append("Elapsed time:")
.append(tm.getElapsedTime())
.append("\n Thread id: ")
.append(tm.getThreadId())
.append("\n Original url: ")
.append(tm.getOriginalFullURL())
.append(")");
for (int i = 0; i < tinfo.length; i++) {
ThreadInfo e = tinfo[i];
el = e.getStackTrace();
if (el != null && el.length != 0) {
for (int n = 0; n < el.length; n++) {
StackTraceElement frame = el[n];
if (frame == null) stackString.append(" null stack frame\n");
else stackString.append(" null stack frame").append(frame.toString()).append("\n");
}
}
}
threadCount++;
}
logger.warn(stackString);
} else {
logger.warn(
"A thread took to long but the system seems to be really clogged so we don't send this one.");
}
}
