Mailing List Archive: conserver: users

memleak?

Index | Next | Previous | View Threaded

wernli at in2p3

Jun 11, 2009, 8:13 AM

Post #1 of 4 (1404 views)
Permalink

memleak?

Hi,

I'm having some memory related problems on linux 2.6 x64.
One of our conserver handles 1700 consoles, most of which are 'exec' types.
1300 use 'ipmitool' wrappers.

Of course there's lots of ipmitools in the process list.
Also, there's 100 conserver processes which use around 8M of RAM each upon
startup. Now these grow over time, reaching huge amounts (over 200M each).
I know there's some ipmitool processes which seem to memleak, and which I
kill on a regular basis. However, this seems to be the case with conserver
processes themselves, and to a bigger extent.

Any ideas on what I might be doing wrong?

_______________________________________________
users mailing list
users [at] conserver
https://www.conserver.com/mailman/listinfo/users

nstraz at redhat

Jun 11, 2009, 8:29 AM

Post #2 of 4 (1302 views)
Permalink

Re: memleak? [In reply to]

On Jun 11 17:13, Fabien Wernli wrote:
> Also, there's 100 conserver processes which use around 8M of RAM each upon
> startup. Now these grow over time, reaching huge amounts (over 200M each).
> I know there's some ipmitool processes which seem to memleak, and which I
> kill on a regular basis. However, this seems to be the case with conserver
> processes themselves, and to a bigger extent.
>
> Any ideas on what I might be doing wrong?

Can you run conserver under valgrind? Perhaps run it on a test box with
only a few ipmi console for a few hours.

Be sure to use --leak-check=full.

Nate
_______________________________________________
users mailing list
users [at] conserver
https://www.conserver.com/mailman/listinfo/users

bryan at conserver

Jun 11, 2009, 4:28 PM

Post #3 of 4 (1311 views)
Permalink

Re: memleak? [In reply to]

On Thu, Jun 11, 2009 at 05:13:34PM +0200, Fabien Wernli wrote:
> I know there's some ipmitool processes which seem to memleak, and which I
> kill on a regular basis. However, this seems to be the case with conserver
> processes themselves, and to a bigger extent.
>
> Any ideas on what I might be doing wrong?

What version are you using? I don't want to assume.

There's also the --with-dmalloc configure option, which I've used in the
past to track down and eliminate leaks. I can't guarantee they're all
gone, but it was an effort (at one point) to remove any I can find).
Those additional hooks might come in handy.

Bloat like this shouldn't happen...so there's definitely a problem
*somewhere*.

Bryan
_______________________________________________
users mailing list
users [at] conserver
https://www.conserver.com/mailman/listinfo/users

wernli at in2p3

Jun 12, 2009, 12:20 AM

Post #4 of 4 (1316 views)
Permalink

Re: Re: memleak? [In reply to]

On Thu, Jun 11, 2009 at 04:28:48PM -0700, Bryan Stansell wrote:
> What version are you using? I don't want to assume.

conserver: conserver.com version 8.1.16
conserver: default access type `r'
conserver: default escape sequence `^Ec'
conserver: default configuration in `/etc/conserver/conserver.cf'
conserver: default password in `/etc/conserver/conserver.passwd'
conserver: default logfile is `/var/log/conserver/server.log'
conserver: default pidfile is `/var/run/conserver/pid'
conserver: default limit is 16 members per group
conserver: default primary port referenced as `conserver'
conserver: default secondary base port referenced as `0'
conserver: options: dmalloc, libwrap, openssl, pam
conserver: dmalloc version: 5.3.0
conserver: openssl version: OpenSSL 0.9.7a Feb 19 2003
conserver: built with `./configure --with-openssl=/var/tmp/openssl-0.9.7a
--with-libwrap --with-dmalloc --with-pam --prefix=/opt
--sysconfdir=/etc/conserver --with-logfile=/var/log/conserver/server.log
--with-pidfile=/var/run/conserver/pid'

> There's also the --with-dmalloc configure option, which I've used in the
> past to track down and eliminate leaks. I can't guarantee they're all
> gone, but it was an effort (at one point) to remove any I can find).
> Those additional hooks might come in handy.
>
> Bloat like this shouldn't happen...so there's definitely a problem
> *somewhere*.

I'll see what I can find out using valgrind, as suggested by Nate

_______________________________________________
users mailing list
users [at] conserver
https://www.conserver.com/mailman/listinfo/users

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads