RE: Performance Tuning - Single request brings server to its knees

From: <jcfolsom_at_pureperfect.com>
Date: Tue, 09 Feb 2010 14:57:15 -0700

Here are some questions:

Is there specific some reason why you are running bulk updates inside rather than outside the container?

At what point during an individual HTTP request does the request hang?

What version of Toplink are you using?

What storage engine is it using for MySQL?

Are you using connection pools?

How is the pool configured?

Do you have any shared state (particularly Toplink) between HTTPRequest threads? If so, is it a source of contention (i.e. locking)?

What percentage of the CPU are you using?

Is the JVM running with concurrent GC enabled?

What is the maximum number of request threads?

Are you sure it isn't just because you are running Windows?

-------- Original Message --------
Subject: Re: Performance Tuning - Single request brings server to its
knees
From: Dhiru Pandey <Dhiru.Pandey@Sun.COM>
Date: Tue, February 09, 2010 4:30 pm
To: users@glassfish.dev.java.net

Hi,

Its hard to guess where the issue would be with out more data like CPU
profiles, jstack dumps for GlassFish.

I would start with looking at the CPU load on then machine when these
bulk uploads/deletes are done. If your CPUs are 100 % pegged at that
time then you are running out of processing resources on this machine -
you should look at moving either the database or GlassFish to its own
dedicated machine.

If the CPUs utilization is not 100%, try to increase the number of
worker threads on the HTTP path - by default this is set to grow
maximally to 5. Try bumping this up. You can do this by using the Admin
Console and by traversing to the following node of the Tree (left pane) :
CommonTask->ThreadPools->http-thread-pool

and then changing the Max. Thread Pool Size to some number greater than
5 (like 10 or 20). You will need to restart the server for these
settings to take effect.

(I am assuming here that the all the users and their interaction with
the application including the bulk update/delete is done using the web)

Hope this helps. Good luck

glassfish@javadesktop.org wrote:
> First off, I'm in the deep end here. I fully acknowledge the noobness of this request but I've looked at a lot of tuning guides and have tried various settings in the admin console all to no avail so I would really appreciate some assistance.
>
> [b]The problem:[/b]
> We have an application deployed that runs great if you have one user on it problem is that we kinda want a few more then that. When the administrators do a bulk upload/delete which is fairly heavy in the number of database transactions it completely hangs for all users and badly. Its almost like there is only one thread and until it's finished nothing else can happen. I'm currently staring at two tabs (one admin/ one just to login) in firefox which have been spinning for the last 10 minutes!
>
> [b]Our Setup[b]
> Server:
> 3.2 ghz Intel XEON
> 2 processors
> 8GB RAM
> Windows 64 bit
>
> Glassfish: v2.1
> DB: MSSQL
> Other: Using Toplink and JSP
>
> [b]Question[b]
> How do I go about narrowing down what the problem is? Like I've said I have tried the tuning tips on several pages but they seem to have no effect. A lot of them refer to using the built in monitoring in the admin panel but I have to be honest I look at the results and have no idea how to work with the results.
>
> Thanks for your time. Please let me know if there is anything else you need.
> [Message sent by forum member 'scomley' (register@comley.org)]
>
> http://forums.java.net/jive/thread.jspa?messageID=385608
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
> For additional commands, e-mail: users-help@glassfish.dev.java.net
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@glassfish.dev.java.net
For additional commands, e-mail: users-help@glassfish.dev.java.net