Is there a limit to mod_jk?

[John Larsen <jo...@javapipe.com>]

Hello,

Is there a limit to the number of workers instances mod_jk can handle?

I currently have 38 tomcat instances running on a machine. I have had up to
75 in the past, but this machine I keep getting bind exception errors but
lsof -i :port comes up empty when that particular tomcat is offline.

Im at loss as to what is causing it. Usually bind exceptions is when
another java process is running on the same port - but not the case here.
Even if i shutdown apache where mod_jk isnt being used tells me this is
really unrelated to mod_jk

mod_jk.log:
[Thu Apr 25 04:14:07.458 2019] [30178:139932601325312] [error]
ajp_service::jk_ajp_common.c (2796): (w314) connecting to tomcat failed
(rc=-3, errors=2, client_errors=0).
[Thu Apr 25 04:14:07.458 2019] [30178:139932601325312] [info]
jk_handler::mod_jk.c (2991): Service error=-3 for worker=w314

I tried updating mod_jk to 1.2.46


John Larsen

Re: Is there a limit to mod_jk?

[Rainer Jung <ra...@kippdata.de>]

Am 25.04.2019 um 06:22 schrieb John Larsen:
> Hello,
> 
> Is there a limit to the number of workers instances mod_jk can handle?

There will be limits due to general file descriptor limits, each TCP 
connection counts as a file descriptor. But I am not aware of a limit on 
the number of workers per se.

> I currently have 38 tomcat instances running on a machine. I have had up to
> 75 in the past, but this machine I keep getting bind exception errors but
> lsof -i :port comes up empty when that particular tomcat is offline.

These numbers do not look that high. A nasty limit on some linuxes is 
the number of processes (maxproc or ulimit -u). It is counted per user 
and despite its name it actually counts tasks, and each thread in each 
process of a user is such a task. So it actually counts the number of 
threads in all processes fo some user id. The first process that hits 
the limit will only observe, that it can not start a new thread. 
Typically this will result in very unexpected behavior.

> Im at loss as to what is causing it. Usually bind exceptions is when
> another java process is running on the same port - but not the case here.
> Even if i shutdown apache where mod_jk isnt being used tells me this is
> really unrelated to mod_jk

How exactly does the exception look like (full stack)?

> mod_jk.log:
> [Thu Apr 25 04:14:07.458 2019] [30178:139932601325312] [error]
> ajp_service::jk_ajp_common.c (2796): (w314) connecting to tomcat failed
> (rc=-3, errors=2, client_errors=0).
> [Thu Apr 25 04:14:07.458 2019] [30178:139932601325312] [info]
> jk_handler::mod_jk.c (2991): Service error=-3 for worker=w314
> 
> I tried updating mod_jk to 1.2.46

There should be additional log lines directly above the snippet you 
showed here. Could you provide a more complete snippet?

Regards,

Rainer


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@tomcat.apache.org
For additional commands, e-mail: users-help@tomcat.apache.org