Virtual Memory Consumption of Cyrus IMAPd

[Andrew Morgan]

Let me know how that works please.  I'm using pam because I pull users
from ldap and the cyrus admin from /etc/passwd, but I'd be interested in
using the ldap mech too.  I'm running Debian stable here, but sasl and
cyrus are compiled from source.

	Andy

On Mon, 19 Jul 2004, Lenny wrote:

> Very useful information. Thank you, Igor.
>
> Lenny
> --
>
> "Wisdom is to a man an infinite Treasure" - Anonymous"
>
>
> Quoting Igor Brezac <igor at ipass.net>:
>
> > On Mon, 19 Jul 2004, Lenny wrote:
> >
> >> Ya, I could do that. Would there be a gain for me to compile the latest from
> >> source and using the ldap mech as opposed to the "-n0" flag suggestion
> >> that was
> >> made?
> >
> > -n0 is very taxing on the OS specially if you have a lot of users.
> > saslauthd/ldap does not leak (so you do not have restart saslauthd
> > every so often) and it is much faster than pam/ldap.
> >
> > --
> > Igor
> >
> >>
> >> Lenny
> >> --
> >>
> >> "Wisdom is to a man an infinite Treasure" - Anonymous"
> >>
> >>
> >> Quoting Igor Brezac <igor at ipass.net>:
> >>
> >> > On Mon, 19 Jul 2004, Lenny wrote:
> >> >
> >> >> Fedora Core 1's saslauthd doesn't seem to like me setting the mech to
> >> >> ldap. This
> >> >> works great on Core 2, but I'm not in a postition to migrate all
> >> the cyrus
> >> >> servers to Fedora Core 2 as of yet. So this is a temporary
> >> problem for me :)
> >> >
> >> > Can you build saslauthd from the source...
> >> >
> >> > -Igor
> >> >
> >> >>
> >> >> Lenny
> >> >> --
> >> >>
> >> >> "Wisdom is to a man an infinite Treasure" - Anonymous"
> >> >>
> >> >>
> >> >> Quoting Igor Brezac <igor at ipass.net>:
> >> >>
> >> >> > pam is known to leak memory on some OSs.  This is really not saslauthd
> >> >> > problem.  Have you tried to use saslauthd built-in ldap mech?
> >> >> >
> >> >> > -Igor
> >> >> >
> >> >> > On Mon, 19 Jul 2004, Lenny wrote:
> >> >> >
> >> >> >> Using the pam mech which in turn is authing against an
> >> openldap server.
> >> >> >>
> >> >> >> I've tried the -n0 flag that was just mentioned in hopes this solves
> >> >> >> the problem
> >> >> >> :)
> >> >> >>
> >> >> >> Lenny
> >> >> >> --
> >> >> >>
> >> >> >> "Wisdom is to a man an infinite Treasure" - Anonymous"
> >> >> >>
> >> >> >>
> >> >> >> Quoting Igor Brezac <igor at ipass.net>:
> >> >> >>
> >> >> >> > What saslauthd authentication mechanism are you using?
> >> >> >> >
> >> >> >> > -Igor
> >> >> >> >
> >> >> >> > On Mon, 19 Jul 2004, Lenny wrote:
> >> >> >> >
> >> >> >> >> Not sure about RHEL, but on Fedora Core 1 the saslauthd process
> >> >> >> eats tons of
> >> >> >> >> swap until eventually there's no memory left. Great fun! This of
> >> >> >> >> course stops
> >> >> >> >> cyrus from being able to process imap and pop logins (if
> >> you're using
> >> >> >> >> saslauthd
> >> >> >> >> for authentication). A simple restart usually fixes this
> >> >> >> particular problem.
> >> >> >> >>
> >> >> >> >> This may or may not be related to what you're seeing, but I figured
> >> >> >> >> I'd throw
> >> >> >> >> this out there.
> >> >> >> >>
> >> >> >> >> Lenny
> >> >> >> >> --
> >> >> >> >>
> >> >> >> >> "Wisdom is to a man an infinite Treasure" - Anonymous"
> >> >> >> >>
> >> >> >> >>
> >> >> >> >> Quoting Mohamed Magdi Abbas <abbasmm at longwood.edu>:
> >> >> >> >>
> >> >> >> >> > I'm running cyrus impad(v 2.2.6), used to run 2.1.16, on
> >> >> RHEL 3.0 and
> >> >> >> >> > noticed that the amount of swap space free is decreasing.
> >> When I was
> >> >> >> >> > running v2.1.16, the swap free would decrease until the
> >> >> kernel oom(out
> >> >> >> >> > of memory) killer would activate and start killing
> >> processes which
> >> >> >> >> > essentially grinds the system to a halt. I doubt that this
> >> >> is normal.
> >> >> >> >> > Could this point to a memory leak in cyrus. Just for note, I
> >> >> have not
> >> >> >> >> > yet experienced this with the latest version, but the
> >> swap free is
> >> >> >> >> > decreasing.
> >> >> >> >> >
> >> >> >> >> > By the way I'm using simons RPMS from invoca.ch. Could someone
> >> >> >> shed some
> >> >> >> >> > light on this. Anything will help...
> >> >> >> >> >
> >> >> >> >> > Mohamed Magdi Abbas
> >> >> >> >> > Systems Engineer
> >> >> >> >> > Longwood University
> >> >> >> >> > abbasmm at longwood.edu
> >> >> >> >> >
> >> >> >> >> > ---
> >> >> >> >> > Cyrus Home Page: http://asg.web.cmu.edu/cyrus
> >> >> >> >> > Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
> >> >> >> >> > List Archives/Info:
> >> http://asg.web.cmu.edu/cyrus/mailing-list.html
> >> >> >> >>
> >> >> >> >>
> >> >> >> >> ---
> >> >> >> >> Cyrus Home Page: http://asg.web.cmu.edu/cyrus
> >> >> >> >> Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
> >> >> >> >> List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
> >> >> >> >>
> >> >> >> >
> >> >> >> > --
> >> >> >> > Igor
> >> >> >>
> >> >> >>
> >> >> >>
> >> >> >
> >> >> > --
> >> >> > Igor
> >> >>
> >> >>
> >> >>
> >> >
> >> > --
> >> > Igor
> >>
> >>
> >>
> >
> > --
> > Igor
>
>
> ---
> Cyrus Home Page: http://asg.web.cmu.edu/cyrus
> Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
> List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
>
---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html