Hi
Very interesting discussion about JVM tuning. I can't help much with
that, but I do have a suggestion for investigating your database.
If you have SQL Server you can run the SQL profiler to watch for slow
queries; I've used this to isolate the queries that were actually slow
(lots of read/writes or cpu, for example) from the queries that were
slow because they were waiting for the other query to finish. This can
be much more effective than looking at the queries on slow pages because
you never really know why the page (or even the query that the page
does) is slow.
SQL profile does add a bit of ovehead so it is not the kind of thing you
want to run "all of the time", but I've run it for extended periods (a
day or two) and learned a lot about that helped improve performance.
Thanks
Mark
-----Original Message-----
From: Rick Root [mailto:rick.r...@webworksllc.com]
Sent: Thursday, February 05, 2009 7:15 PM
To: cf-talk
Subject: Re: JVM Tuning and Garbage Collection
I've taken advice from several of you and restarted my services... so
far so good.
I lowed the memory on each instance to 768m - min and max - and added
the garbage collection interval. I also increased the maxpermsize
java.args=-server -Xmx768m -Xms768m -Dmail.host=www.classcreator.com
-Dsun.io.useCanonCaches=false -Dsun.rmi.dgc.client.gcInterval=600000
-Dsun.rmi.dgc.server.gcInterval=600000 -XX:MaxPermSize=256m
-XX:+UseConcMarkSweepGC -Dcoldfusion.rootDir={application.home}/
Oops.. I just noticed that I accidentally left the GC method to
UseConcMarkSweepGC .. I didn't intend for that as I switched it to that
last friday night, and monday when activity picked up I was in hell...
switched back to UseParallelGC Tuesday morning, and things smoothed out
considerably.
So far, things are running pretty smoothly. Garbage collection does not
seem to be causing any problems. I am still seeing these weird,
rhythmic activity spikes where every 13-14 minutes, my request activity
graph skyrockets to 20 active requests, while it usually hovers around
2-3 active requests (while processing anywhere from 5-10 requests per
second)
NEXT STEP: I'm going to upgrade from the JVM that shipped with CF 8 to
the latest 1.6.0_12
At the same time as these request spikes, I see JDBC time spikes. A
significant spike of requests could slow down the database server, so
this makes sense to me. Alternatively, a temporary slowdown on the
database server could also lead to a slowdown of request processing
which would lead to the "piling up" of requests.
HOWEVER, I haven't seen anything particularly unusual on the database
server that would cause these rhythmic spikes in activity. No scheduled
tasks running every 13 minutes, etc.
I am having Cacti installed so we can better monitor CPU and disk
activity on our servers so hopefully that will help me problem solve!
Anyway, thanks all, I appreciate your input.
Rick
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
Adobe® ColdFusion® 8 software 8 is the most important and dramatic release to
date
Get the Free Trial
http://ad.doubleclick.net/clk;207172674;29440083;f
Archive:
http://www.houseoffusion.com/groups/cf-talk/message.cfm/messageid:318965
Subscription: http://www.houseoffusion.com/groups/cf-talk/subscribe.cfm
Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4
