freebsd stable - Dec 2004 - 5.3 and maxproc exceeded

Hi,
I'm trying to dig up a problem. A server I have running 201 jails with a
total of ~2200 processes gets about every five minutes:

"maxproc limit exceeded by uid 2, please see tuning(7) and
login.conf(5)"

First of all, uid 2 has only a few sendmail processes running (postfix
alias for sendmail) Second of all, I've upped both kern.maxproc and
kern.maxprocperuid to 32k and 16k, well above what this server actually
runs of total processes.

To make things even more interesting, java dies by signal 6 (Abort)
frequently (it's running Tomcat 4.1) (while I was writing this mail, make
buildworld Abort'ed as well :-I ) and I've had reports of this server
panicing a few times the last 24 hours, but since I haven't been around
when it's happened and no-one noted down the panic messages because they
were too stressed about getting the computer up, I have no clue what might
have happened there.

But if someone have any ideas or guesses about the maxproc problem, I bet
that would help me nesting up the problem. :)

Cheers

        Nik

On Thu, Dec 23, 2004 at 11:49:50AM +0100, niklas@registrar.no
wrote:
> Hi,
> I'm trying to dig up a problem. A server I have running 201 jails with
a
> total of ~2200 processes gets about every five minutes:
> 
> "maxproc limit exceeded by uid 2, please see tuning(7) and
login.conf(5)"
> 
> First of all, uid 2 has only a few sendmail processes running (postfix
> alias for sendmail) Second of all, I've upped both kern.maxproc and
> kern.maxprocperuid to 32k and 16k, well above what this server actually
> runs of total processes.
It may only have a few processes running, but uid 2 was the owner of
the one process that caused the total to exceed the maxproc limit.
> To make things even more interesting, java dies by signal 6 (Abort)
> frequently (it's running Tomcat 4.1) (while I was writing this mail,
make
> buildworld Abort'ed as well :-I ) and I've had reports of this
server
> panicing a few times the last 24 hours, but since I haven't been around
> when it's happened and no-one noted down the panic messages because
they
> were too stressed about getting the computer up, I have no clue what might
> have happened there.
This may also be a side-effect of a resource limit being exceeded.

You should enable crashdumps, so that you can dissect the panic after
the computer reboots.

Kris
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 187 bytes
Desc: not available
Url :
http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20041223/afccaf1d/attachment.bin