cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

Community Tip - Need help navigating or using the PTC Community? Contact the community team. X

Tomcat Tuning in 10.1

MikeIbosh
7-Bedrock

Tomcat Tuning in 10.1

All,
I have a question about the imbedded Tomcat process in Windchill 10.1 M010 MethodServer. Are there separate tuning parameters for the Tomcat process that is imbedded into each MethodServer, or does the Tomcat process share the min/max memory heap with the MethodServer process. If there are separate tuning parameters, where are they located. I know that the WCA is supposed to do this for you, but I can't find where these settings are. Anybody have any idea where to look?
Thanks ...
Mike
8 REPLIES 8
jessh
12-Amethyst
(To:MikeIbosh)

There is no separate Tomcat process in 10.0 and higher -- it's part of
the method server process.

Thus there are no separate tuning parameters for a Tomcat process in
10.0 and higher.

There are some tuning parameters for Tomcat's thread pools, etc, in
WT_HOME/tomcat/config.properties (propagated by executing "ant -f
config.xml reconfigure" from WT_HOME/tomcat). There is rarely a need to
change these parameters, however.

If one is configuring /extremely /large method servers, such that each
is expected to serve well over 300 simultaneous requests (not
simultaneous users but rather simultaneous requests) then one might need
to increase 'ajpMaxThreads'. That's only in the case of an extremely
large method servers, though, since the default is 320. Also, this
setting only has an impact if you also increase 'ajpMaxConn' in Apache's
config.properties (also propagated by "ant -f config.xml reconfigure",
but from the Apache directory). 'ajpMaxConn' defaults to a much lower
value than 'ajpMaxThreads', so 'ajpMaxConn' is actually the throttle
out-of-the-box. You'll know if you exceed the limit imposed by
'ajpMaxConn' if your users start getting 503 errors during heavy load.
Note that Tomcat will misbehave unless (ajpMaxConn)*(# of Apache worker
processes) + 2 is less 'ajpMaxThreads', so when adjusting 'ajpMaxConn'
you have to adjust 'ajpMaxThreads' accordingly if you're increasing
'ajpMaxConn' by a large amount.

Typically the only Tomcat config.properties entry you should need to
adjust is 'mode' -- and that's only for developers. The value of 'mode'
defaults to 'prod', which is short for "production". When developing,
however, one often will want to set 'mode' to 'dev', which is short for
"development". The development mode has lower performance but ensures
that Tomcat will check for changes to JSPs each time a JSP is requested,
whereas in production mode such checks are not performed once a JSP is
loaded.

--
Jess Holle

Jess,
Thanks for the information, I'm running 10.1 M010 w/four MethodServers and a BackgroundMethodServer (configured at 4 GB max heap) but my users are still complaining about slow workspace refresh and update times and I wasn't quite sure about tuning. I was looking in the Specialized Admin Guide for possible performance tuning suggestions. Is there a good manual to look at for this type of thing or settings for the amount of memory allocated for the embedded Tomcat process?
Mike

Mike,

Check the view they are using in the workspace, we had an issue where we
had the workspaces using the compare status column and it really slowed
down our users, they now have a view with basic info and one they use
prior to check in with the compare in it - Helped us a lot

Worth a look


Best Regards

Chris Collinson
CAD Administrator
jessh
12-Amethyst
(To:MikeIbosh)

Given your description, I'd say that this has nothing to do with Tomcat
tuning parameters.

There are, however, a lot of things that might need tuning. As to which
it is, I can't hazard a guess from this limited information. This would
be something to engage technical support on.

TomU
23-Emerald IV
(To:MikeIbosh)

We were having significant performance issues in two cases.

The first case was with the embedded browser. If a workspace had more than ~330 items in it, and any of the columns "Compare status", "Out of date", or "Out of date with workspace configuration" were shown in the view, the workspace would take one minute and three seconds (63 seconds) to refresh. If the number of items in the workspace was reduced by even one (below some magic, slightly variable threshold), then the refresh would take 3 seconds. This was extremely repeatable and reproducible across computers and browser versions and types.

The second case was extremely slow refresh times when browsing the Windchill folders from a standalone browser if there was a significant number of objects in the folder. The thumbnails would take forever to show up, even after the page had apparently completed loading.

After weeks of waiting, many calls and WebEX sessions, and sending PTC support many different log files, I finally found the solution on my own. (Packet sniffers are wonderful.) The issue had to do entirely with the antivirus software running on each CAD workstation. By adding an exception to the antivirus program for the Windchill server's URL, both problems were immediately eliminated. Of course after figuring this out, PTC dug up a best practice guide that suggested the same thing - don't use antivirus software on the Windchill server, and don't allow local client antivirus programs to scan the Windchill server (or data sent from it).

Might be something worth looking into.

Tom U.

Tom,
WOW, I have almost the same situation here.
Our workspaces usually contain some 2000-5000 objects, and are very slow to update or just display. We do display the Out Of Date and Modified Locally columns in the client workspace views.
The folder browsing issue is the same as you stated below.
I was able to convince our IT department to exclude the entire Windchill install folder from the virus scans on the server, but I didn’t think of doing the same for the clients. I’ll see if I can convince the IT department to push that out also.
We just upgraded from 9.1 M060 to 10.1 M010 two weeks ago, so I’m still trying to get the performance back to what is was in 9.1.
Thank you very much for the insight !!

Mike Ibosh
jkent
12-Amethyst
(To:MikeIbosh)

Same here. We white listed our Windchill server IP address since it was scanning the traffic between the server and client. We also white listed xtop.exe since the anti-virus was scanning the disk activity of xtop.

Joe Kent
Engineering Systems Administrator
R&D/Leverage™
Structural Brand Development / Mold Manufacturing
"With Us, Ideas Take Shape"

Tel: 816-525-0353 x6527

www.rdleverage.com<">http://www.rdleverage.com>

Get Access To… Packaging That Sells - SPE Blow Molding Conference - Pack-Expo International –
And more by visiting here -
TomU
23-Emerald IV
(To:MikeIbosh)

Here are the comments from PTC tech support.
Announcements


Top Tags