I reported the same problem (OOM's on a 5GB system which should have had 3-4 GB free). It's a pure kernel problem and hopefully the new U1 kernel fixes this issue - it's in the redhat bugzilla, to quote:

[this is the 2.6.9-5.0.5ELsmp kernel]

OK ... there are known memory leak issues on that kernel

Your problem is similar, but not exactly like:

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=150971

we'll see whether the new U1 kernel helps... so long the situation seems much better... Cheers, MaZe.

On Friday 10 June 2005 07:29 pm, israel.garcia@cimex.com.cu wrote:

List, I have the two servers CentOS 4.0 (running postfix, mailscanner) on this hardware: HP netserver 2000, PIII - 1GHz, 512 ram, two Hdd SCSI - 36 GB (RAID 1), I have monitored this server for a couple of months and I've seen that the memory used is increasing every week until it reaches all available memory and the server crashes.... at this point mailscanner crashes or sometimes the server gets freezed....my question is?

Is this a memory leak?

I 've always used this combination (mailscanner, postfix on Whitebox) and never got a problem, but when I upgrade my system to CentOS (kernel 2.6) this problem began..

Are you running spamassassin?  I had the same problem after upgrading to v3 of sa which rh4/c4 ships with.

can you help me?

regards

Israel