Hi
Some times my server grizzly-comet-webserver-1.9.17-SNAPSHOT gets stuck
while processing the request.
Test Case Scanario:
The client i'm using is the sample application from the trunk, comet-chat. I
open a browser window(index.html) and then keep on hitting the refresh
button in the browser. The browser hangs(waiting for resource from server)
after some hits. Pls find the screenshot of browser.
http://www.nabble.com/file/p24033280/stuck_request.jpg stuck_request.jpg
I i'm using the latest trunk code (1.9.17-SNAPSHOT, rev 3321). I have tested
on my desktop(win XP 32 bit single core) and server class machine (4 cpu,64
bit, linux). The chances of reproducing the issue in SERVER is MORE than in
my desktop.
Please find the attached thread dump when the server is stuck.
http://www.nabble.com/file/p24033280/stuck_threaddump.txt
stuck_threaddump.txt .
One (strange??) thing I have noticed is that if I put a sleep in as follows
I'm not able to reproduce this issue.
method DefaultThreadPool.execute(Runnable){
....
if (running){
try {
Thread.sleep(5);
} catch {....}
workQueue.offer(task);
queueSize.incrementAndGet();
}
}
To debug I have added some sysout statements in the following parts.
method DefaultThreadPool.execute(Runnable){
....
if (running){
System.out.println("DefaultThreadPool.execute()"+" ,
"+System.nanoTime()+","+Thread.currentThread().getName()
+ " , Offer , "+ task);
workQueue.offer(task);
queueSize.incrementAndGet();
}
}
method FixedThreadPool.BasicWorker.dowork(){
.......
Runnable r = getTask();
System.out.println("FixedThreadPool.BasicWorker.dowork()"+" ,
"+System.nanoTime()+","+Thread.currentThread().getName()+" , Take , "+r);
if (r == poison || r == null){
return;
}
......
}
Please find the sysout messages both in the normal case(not stuck) and stuck
case. I have the results in csv format.
Normal(not stuck) sysout messages :
http://www.nabble.com/file/p24033280/normal.csv normal.csv
Stuck sysout Messages :
http://www.nabble.com/file/p24033280/stuck.csv
stuck.csv
Here the number of invocations of the workQueue.take() and workQueue.offer()
in a stuck request will be less than in the case of normal request.
Kindly Look into this.
Please let me know if any other details are required to reproduce this
issue.
Thanks,
Abey Tom.
--
View this message in context: http://www.nabble.com/grizzly-comet-webserver-gets-stuck-sometime-while-processing-request-tp24033280p24033280.html
Sent from the Grizzly - Users mailing list archive at Nabble.com.
