tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From peter lin <peter....@labs.gte.com>
Subject Re: excessive cpu use with 4.0.4
Date Tue, 01 Oct 2002 11:44:20 GMT

I don't know if others have mentioned this, but have you tried using
other vm's?

I've done several benchmarks and stress tests with 4.x and haven't seen
any problems with a real webapp. The webapp is fairly complicated with a
couple hundred jsp pages, so arguably a webapp that uses JSP
would/should expose load problems faster than servlet webapps.

peter

Seb B wrote:
> 
> Hi,
> 
> I'm running a stand-alone Tomcat 4.0.4 (not using jsp, so no need
> to upgrade to 4.0.5) on Linux RedHat 7.3, with IBM's 1.3.1 jvm.
> There is very little static content to serve and the server
> doesn't have to listen on port 80, so I haven't tried learning
> how to configure it behind Apache.
> 
> Everything works great in the lab, so I'm now experimenting in
> our production system (whose servers are running other servlet
> containers so far), behind a load balancer.
> 
> On the last run, 4 of the java threads started taking all of the
> cpu. I stopped sending traffic to tomcat, and those 4 went on
> taking all the cpu they could (for several days). So it looked
> like they were stuck in some infinite loop somewhere. Of course,
> I suspect this would come from our code. I tried enabling all
> sorts of logging in our code, but couldn't get anything from the
> process.
> 
> As a last resort, I sent the process a "SEGV" signal, which leads
> the jvm to print stack traces. All 4 threads (I knew their PID
> from ps/top) had the following stack trace:
>     "HttpProcessor[9090][44]" (TID:0x100502B8,
> sys_thread_t:0x46DCDD98, state:R, native ID:0x10442) prio=5
>         at
> org.apache.catalina.connector.http.HttpProcessor.run(HttpProcesso
> r.java:1125)
>         at java.lang.Thread.run(Thread.java:512)
>          PID: 8013
> None of the other HttpProcessor threads were on that same line
> (1125), all those I looked at were (as they should be I guess):
>     "HttpProcessor[9090][43]" (TID:0x10050300,
> sys_thread_t:0x46DCD918, state:CW, native ID:0x10041) prio=5
>         at java.lang.Object.wait(Native Method)
>         at java.lang.Object.wait(Object.java(Compiled Code))
>         at
> org.apache.catalina.connector.http.HttpProcessor.await(HttpProces
> sor.java(Compiled Code))
>         at
> org.apache.catalina.connector.http.HttpProcessor.run(HttpProcesso
> r.java:1119)
>         at java.lang.Thread.run(Thread.java:512)
>          PID: 8012
> So *if* the stack traces from the jvm are worthwhile, this would
> point to a tomcat problem.
> 
> I'm posting to this list and not the developer list, as I'm
> wondering whether anyone has experienced such a problem before.
> This has happened in only one of my test runs, I don't know how
> reproduceable it will be.
> 
> Also, this has been asked already, but I didn't see many answers:
> is there any sort of real-world benchmark of what load I can
> expect in such a set-up ? I understand it depends a lot on what
> the servlets do, so I'm wondering how big the biggest sites using
> tomcat are; John Turner mentioned >500k for his, is anyone doing
> anything bigger ?
> 
> Thanks a lot,
> 
> Seb
> 
> --
> To unsubscribe, e-mail:   <mailto:tomcat-user-unsubscribe@jakarta.apache.org>
> For additional commands, e-mail: <mailto:tomcat-user-help@jakarta.apache.org>

--
To unsubscribe, e-mail:   <mailto:tomcat-user-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:tomcat-user-help@jakarta.apache.org>


Mime
View raw message