Tomcat/Cocoon on Linux woes

[Matthew Langham <ml...@s-und-n.de>]

(Posted this yesterday from another email-address but it hasn't made it thru
yet)

--------------------------------------------------------
Help: Tomcat 3.3 with Cocoon 2.0 hangs(?) on SuSE Linux!
--------------------------------------------------------

We are currently experiencing a PITA problem on several SuSE Linux systems
(with different kernel versions) where Tomcat 3.3 will just seem to hang
after a while. I am hoping someone out there has perhaps an idea what we
need to examine to find out what is going on.

We are using WGET to send parallel requests into Cocoon 2.0. In all there
are about 4 - 6 requests being processed in parallel at any one time. The
actual pipeline that is called fetches XML data from an external source and
then uses the Cocoon caching mechanism to keep the data in the cache.

The systems will run ok for various lengths of time (we have seen 2000-3000
requests be processed before the hang - but also seen 17000 requests before
the same thing happens) before basically just stoping. There are no errors
to be found in any of the logs (tomcat, Cocoon, console). Tomcat just does
not respond. When looking at the vital signs of the system - what we see
does not really help us to see what is going on.

We are currently checking Cocoon and our additional components for memory
leaks (which would seem to be the most obvious cause) but if this is the
case shouldn't we be seeing OutOfMemory Errors somewhere? We have also
backported fixes that were made in the StoreJanitor stuff in later Cocoon
versions. Using TOP to check on the processes seems to indicate a growth in
the number of Java threads - but this is may be caused by the incoming
requests not being processed. So we are not sure whether this is the cause -
or just a result of the hang.

What we have found is that reducing the number of file handles will cause
the hang to happen sooner. This however may be a separate thing and have
nothing to do with the original problem. Of course if file handles are being
leaked somewhere then we would still need to find out where that problem is
coming from.

We are using Sun JDK 1.3.1, Tomcat 3.3.

Ok, so here are my questions to you Linux/Cocoon gurus out there:

Have you seen anything like this happen - and if so what caused it?
How many file handles do you have set on your Linux system? Do you have some
pointers to the number we should have?

We are also considering running the same application in Resin to see if this
prompts at least some error message. What do you think?

It really does seem as though the application is eating something up and
then running out of whatever it is - but it is just not telling us :(.

Sorry this is so long but I am hoping some of the information is setting off
an alarm-bell or two out there.

Any tips, ideas welcome.

Thanks

Matthew Langham

--
Open Source Group               sunShine - Lighting up e:Business
=================================================================
Matthew Langham, S&N AG, Klingenderstrasse 5, D-33100 Paderborn
Tel:+49-5251-1581-30  mlangham@s-und-n.de - http://www.s-und-n.de
           Weblogging at: http://www.need-a-cake.com
=================================================================



---------------------------------------------------------------------
Please check that your question has not already been answered in the
FAQ before posting. <http://xml.apache.org/cocoon/faqs.html>

To unsubscribe, e-mail: <co...@xml.apache.org>
For additional commands, e-mail: <co...@xml.apache.org>

Re: Tomcat/Cocoon on Linux woes

[Michael Hartle <mh...@hartle-klug.com>]

Matthew Langham wrote:

>What we have found is that reducing the number of file handles will cause
>the hang to happen sooner. This however may be a separate thing and have
>nothing to do with the original problem. Of course if file handles are being
>leaked somewhere then we would still need to find out where that problem is
>coming from.
>
>We are using Sun JDK 1.3.1, Tomcat 3.3.
>
Which versions of SuSE Linux did you try ?

We have a test development & presentation system up and running using 
SuSE Linux 7.2 Professional, Suns JDK 1.4.0 and Tomcat 4.0.1 that shows 
no such issues. When we tested several different configurations, we 
found that Suns JDK 1.3.1 would not run properly, but segfault often on 
our system (P4 1,5Ghz/512MB); switching to Blackdowns JDK 1.3.1 solved 
this issue completely.

Then we found that the JDK1.3.1 still suffers from many VM-slowing or 
-killing bugs like a network/garbage-collector issue ("many parallel 
connections" definitely ring a bell here), a multithreading JPEG encoder 
bug and others. After switching to Suns JDK 1.4.0, these bugs seem to 
have disappeared and we are quite confident regarding the stability of 
the system.

Best regards,

Michael Hartle,
Hartle & Klug GbR



---------------------------------------------------------------------
To unsubscribe, e-mail: cocoon-dev-unsubscribe@xml.apache.org
For additional commands, email: cocoon-dev-help@xml.apache.org