We've only got 5.2 apps runing on Jetty, but hot redeploy works great
there and is even without the PermGen leak that was present in Tapestry 4.
Anyway, here's how I would try to diagnose the problem:
1. Trigger the problem so that you have something consuming 100% CPU
2. Find the native ID of the thread that is eating all your CPU power:
$ top -H -u tomcat (or whatever user is running Tomcat)
-H will show individual *threads* and you should be able to find the
offending thread. Remember its ID.
(Hit 'P' to sort by CPU usage)
3. Find Tomcat's PID:
$ sudo jps
4. Get a thread dump from Tomcat:
$ sudo kill -QUIT <TOMCAT_PID>
The thread dump will be written to Tomcat's stderr, wherever that is
redirected
5. Inspect the thread dump
Convert the thread ID from step 2 to hexadecimal and look for it in the
thread dump from step 4. The thread you're looking for will say "nid=xxx",
where xxx is the offending thread's native ID written in hexadecimal. The
thread's stacktrace should give you a clue as to what's going on,
especially if you get more samples to see where the thread is spending
most of its time.
You can also hook up VisualVM to Tomcat and inspect threads from there.
There's a thread inspector plugin in the plugin repository.
Hope that helps!
Martin
On Wed, 20 Jul 2011 11:16:24 +0200, <p.stavrini...@albourne.com> wrote:
Hi All,
This is a tricky question one since I do realize there are several
possibilities to explain the problem, not all of which are Tapestry
related. We have battled to understand what is going on in production
for the past year or so, and have tried to pinpoint where the problem
lies and come up short. If anyone has experienced anything similar
*please share.
Our environment is:
Ubuntu Server
Tapestry 5.1
JDK 6
Tomcat 6
Apache in front (reverse proxy)
We are deploying our Tapestry apps to Tomcat 6, and they do run great
even with load, and it remains very stable as the session count
increases; this is when a server restart is done after a deployment.
The problem appears to be specific to 'Tapestry applications' and hot
deployment. We run several apps, both small and large and they all react
the same way... If we stop then undeploy any of these Tapestry apps, and
afterwards do a hot deploy with a newer version (without the restart),
the CPU usage shoots to 100% and stays there. JSP based apps pure
Servlets don't appear to react in the same way.
Has anybody experienced something similar?
Kind regards,
Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tapestry.apache.org
For additional commands, e-mail: users-h...@tapestry.apache.org
