[Orca-dev] Memory leak

[Christophe Kalt]

On Jan 11, Dmitry Berezin wrote:
| This problem was reported before. 
| We run orcallator.se with SE 3.4 on Solaris 8 and I noticed that SE process
| increases its memory footprint with time. Unfortunately, at this time I do
| not know if the memory leak is in SE or orcallator script, and so far have
| not had time to investigate this. At first glance, it appears that the
| amount of "leaking" memory depends on the number of disks on the system (I
| will try to confirm this shortly), so it looks like a good place to start
| digging...
| Any information on this is greatly appreciated.

Alright..  I'm happy to help although my time is limited.  I'm
running various SE scripts in parallel right now to try and
narrow things down a bit.

Not sure about disks, most of my systems are similar and i
don't have reliable data for most of the ones that aren't
right now.

I restarted the SE processes on all my boxes yesterday, and
identical systems have a slightly different memory footprint
right now.  For example:
21352 vs 22280 for identical systems
22584 for a system with twice as many CPUs and bigger drives
(but same number of drives).
23848 for a system which has more drives.

Bigger footprints for systems with more partitions though, it
seems.  But hard to tell whether it's different because of
excessive leakage or just because there simply is more data.
I wish i had recorded all the footprints before restarting
yeserday.