tomcat-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rainer Jung <>
Subject Re: mod_jk errors and http status code 503 under load
Date Thu, 06 Sep 2007 14:42:21 GMT
Damien Corbishley wrote:
> Hi,
> I'm getting stuck with an error we are seeing on our production
> server under load.
> We upgraded our combination of Apache,Tomcat,Java all running on
> Solaris 10 to
> Apache 2_2.0.55.build2

This doesn't sound like a real Apache httpd version. You can check the 
real version with "httpd -V" or when using log level info in the error 
log of httpd.

Worker or prefork MPM?

> Tomcat 5.5.23
> Java 1.5.0_06-b05

If you are in a general update situation, switch to 1.5.0_12 or whatever 
seems to be recent and not only a couple days old.

And finally mod_jk-version?

> Our Apache Tomcat connector is AJP/1.3.

What's the configuration of the connector element in server.xml?

> It's all installed on the same server, we have no tomcat clusters and
> no load balancers.
> With a few (10) concurrent users accessing the various JSP's
> everything worked as expected, all our functional tests passed.
> We opened more up live traffic to the server and starting seeing the 
> following error message:
> [Thu Sep 06 02:23:00 2007] [error] ajp_service::jk_ajp_common.c
> (1758): Error connecting to tomcat. Tomcat is probably not started or
> is listening on the wrong port. worker=worker1 failed

Is Tomcat on the same machine, or on another one?

What do you get for "netstat -an| grep TOMCATPORT" where you replace 
TOMCATPORT with the port number your Tomcat AJP connector listens on?

How many ESTABLISHED, how many of other states (SYN_SENT, CLOSE_WAIT etc.)?

What's you mod_jk configuration?

> On the client side this was being reported as a 503 and our 
> "intelligent" browser was re-submitting the request and generally,
> on the second time, the request would be successful.
> After an hour we then had a Solaris panic, and after analyzing the
> crash using scat I can see that the cause was:

Solaris panic looks very bad. I would definitely open a bug with sun. 
I'm not saying that the original problem is there, but it is very 
unusual, that the whole system panics.

> panic string:   BAD TRAP: type=9 rp=2a1015332d0 addr=0 mmu_fsr=0
> and looking further into the trap, it seems to be network related.
> <trap>ip:tcp_conn_request+0x894 (0x4, 0x2000, 0x136ec00,
> 0x3a2b9a44040, 0x40e03960300, 0x40dedc247c0)
> ip:squeue_drain+0xe4 (0x300015ddec0, 0x40e2021c1c0, 0x411666de, 
> 0x135f2c0, 0x3a2b9a44040, 0x3)
> ip:squeue_enter+0x358 (0x300015ddec0, 0x38fca5e0d80, 0x3000426b640, 
> 0x3a2bd674c00, 0x0, 0x1)
> ip:tcp_wput - frame recycled
> I am still trying to work out if the crash is a symptom of the
> browsers resubmitting the failed requests and we've hit a Solaris bug
> that we haven't got a patch for,
> Or if the two errors are unrelated and it was coincidence that we 
> panicked whilst having mod_jk errors.
> Any help/pointers in resolving the mod_jk error would be appreciated,
> I included the brief crash information as it may be relevant.
> Thanks
> Damien



To start a new topic, e-mail:
To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message