Hey

We came to the conclusion through rigorous testing and monitoring of stats
etc over the past 6 months.  We used JVMStat to get the data we required
from the JVM from this we adjusted the JVM configs etc and we used SQLDiag
to get info from SQL Server at the time of a hung server so we knew that SQL
Server was pretty busy and indeed we knew that there were exclusive locks on
some core tables...(when we say hung we mean a CF server which has literally
just stopped serving pages mainly due to queued up requests). We have some
serious logging of EVERYTHING which is going on each webserver at any given
time.

You should only be using ParallelGC if you have more than 1 CPU on board the
box, MarkSweep will have serious impact and performance issues and could
crash the server if you have only 1 CPU and you have a pretty busy server.

Are you running IIS?  If you are try running an iisreset.exe on the IIS
Server instead of rebooting or restarting as this will severe the connection
to SQL Server and start afresh.  We are probably going to buy and upgrade
our drivers to DataDirect 3.4 or move totally over to Microsoft JDBC Type 4
as I think the 3.3 driver which ships with CFMX is faulty under load in its
vanilla CF config.

Any help I can provide OL I will be glad to help.

N




-----Original Message-----
From: Bert Dawson
To: CF-Server
Sent: 10/01/2005 19:41
Subject: Re: stability

We've set -Xms862m -Xmx862m (and both to 512 on the box withonly a gig
of ram), so i don't think it can be the 1.8gb thing. (the 1.4 gig
usage i mentioned was reported in task manager).

As for garabge collection, we've changed from (i think):
-XX:+UseParallelGC 
to:
-XX:+UseConcMarkSweepGC 

Prior to that switch then JRun would just crash, and require a manual
restart, and now it will re-start itself, which i suppose is a
blessing.

The main question is where to go next?

Cheers
Bert

On Mon, 10 Jan 2005 20:34:37 +0200, Erki Esken <[EMAIL PROTECTED]>
wrote:
> On Mon, 10 Jan 2005 17:41:03 +0000, Bert Dawson <[EMAIL PROTECTED]>
wrote:
> > I've got a couple of servers running enterprise and they keep
crashing.
> > Usually they restart themselves, and the only downside is everyone
> > losing sessions, and delayMS and CPU go through the roof for a
minute
> > or two.
> > Looking at memory usage and CPU they seem to be fine: CPU around
> > 10-20%, using 1.4gb memory being used on a 4gb box.
> 
> Maybe it's related to the 1.8 GB memory limit of CF? What's the upper
> limit of JVM memory usage on that server? Maybe it dies when memory
> usage goes past 1.8 GB. See also:
>
http://www.macromedia.com/cfusion/knowledgebase/index.cfm?event=view&ext
id=tn_19359
> 
> And try searching "jvm" or "garbage collection" at
> http://fullasagoog.com/ for hints on troubleshooting JVM GC issues.
> There are good blog posts about this.
> 
> --
> Erki
> 
> 



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
Discover CFTicket - The leading ColdFusion Help Desk and Trouble 
Ticket application

http://www.houseoffusion.com/banners/view.cfm?bannerid=48

Message: http://www.houseoffusion.com/lists.cfm/link=i:10:5094
Archives: http://www.houseoffusion.com/cf_lists/threads.cfm/10
Subscription: http://www.houseoffusion.com/lists.cfm/link=s:10
Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.10
Donations & Support: http://www.houseoffusion.com/tiny.cfm/54

Reply via email to