[kronolith] Enabling Kronolith causes huge httpds

[Janne Peltonen]

Hi!

We upgraded into Horde Groupware Webmail Edition 1.1.1 (with Kronolith
H3 (2.2)). As we ran into database performance trouble, we inactivated
everything but IMP and Turba from horde/config/registry.php. Having
found a way around the database problems, we would now like to enable
Kronolith again. However, trying to enable it results in positivele huge
Apache processes (abt 300 megabytes in-core /each/), and we run out of
memory faster than you can say 'cat'... (or, when limiting the number of
concurrent Apache processes so that our 6 gigs of memory doesn't run
out, the service gets unusable, because there aren't enough Apaches to
serve the users). Normally, the Apache process size varies between 35
and 140 megs in-core.

Oh yes, /every/ Apache process gets into at least 285 megs of size in a
couple seconds (a couple new logins?) after enabling Kronolith.

Would anybody have any ideas? We've been unable to reproduce the huge
processes in our test environment, even when using the exact same
database (which is way baffling) - the processes in the test environment
don't seem to get bigger than, say, 90 megs in-core, even with Kronolith.

The test environment is on 32 bit real hardware, production, a 64
bit xen virtual machine (or actually, three of them). There are abt 300
concurrent users on each production server, with abt 40 000 users in
total (that is, not in at the same time). There were some 30 logins /
minute / server this morning when we tried to enable Kronolith once
again (by just editing horde/config/registry.php) and got the huge
Apaches.


Thanks,

Janne Peltonen
IT dept
University of Helsinki
-- 
Janne Peltonen <janne.peltonen at helsinki.fi> PGP Key ID: 0x9CFAC88B
Please consider membership of the Hospitality Club (http://www.hospitalityclub.org)