Bug 799685

Summary: oom-killer activates when plenty of RAM is still available
Product: [Fedora] Fedora Reporter: Slawomir Czarko <slawomir.czarko>
Component: kernelAssignee: Kernel Maintainer List <kernel-maint>
Status: CLOSED CANTFIX QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 15CC: gansalmon, itamar, jonathan, kernel-maint, madhu.chinakonda
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-03-14 14:43:30 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
Section of /var/log/messages
none
dmesg output
none
/var/log/messages none

Description Slawomir Czarko 2012-03-04 09:41:31 UTC 
Description of problem:

After updating to kernel 2.6.42.7-1.fc15.i686.PAE I'm getting oom-killer activated every few hours. I think this might be related to the bug 745241 or 735946 since I've been using fuse and rsync during that time (regression?).


Version-Release number of selected component (if applicable):
2.6.42.7-1.fc15.i686.PAE

How reproducible:


Steps to Reproduce:
1.
2.
3.
  
Actual results:


Expected results:


Additional info:
Comment 1 Dave Jones 2012-03-05 02:48:47 UTC 
we can't do anything with this report without seeing logs from the oomkiller.
Comment 2 Slawomir Czarko 2012-03-05 08:52:10 UTC 
Created attachment 567529 [details]
Section of /var/log/messages

This shows what was logged to /var/log/messages when oom-killer activated.

Do you need any other logs?
Comment 3 Slawomir Czarko 2012-03-14 12:43:25 UTC 
Just got this again now with 2.6.41.10-3 PAE kernel. Attaching dmesg and /var/log/messages
Comment 4 Slawomir Czarko 2012-03-14 12:44:06 UTC 
Created attachment 569971 [details]
dmesg output
Comment 5 Slawomir Czarko 2012-03-14 12:44:36 UTC 
Created attachment 569972 [details]
/var/log/messages
Comment 6 Slawomir Czarko 2012-03-14 12:46:34 UTC 
Machine has 16 GB RAM and there's a few GB always free (or used by filesystem cache). Swap is either not used or barely (less than 100MB).
Comment 7 Slawomir Czarko 2012-03-14 12:47:25 UTC 
Any suggestions on how to debug this?
Comment 8 Slawomir Czarko 2012-03-14 13:13:57 UTC 
Is this due to PAE kernel?

Do you think changing value for vm.lowmem_reserve_ratio will help?
Comment 9 Dave Jones 2012-03-14 14:43:30 UTC 
the real answer is to move to 64bit. Past 8GB, 32bit gets complicated because of so much low memory being used up by page tables. Once fragmentation starts to happen, this sort of thing is to be expected.

You might have memory free when the oom is triggering, but the problem is that you don't have any of the right 'type' of memory (the gfp mask) free.

You might make it perform a bit better by tweaking syscalls, but you're fundamentally in a losing battle just because you're pushing up against the architectural limits.