Re: Why was oom-killer invoked?

tim truman wrote:
Please help me understand why oom-killer was invoked?

You already know that the reason is that your system was out of
memory. Why even ask.

The OOM killer is a terrible thing. I have ranted about it often.
Turning it off is the best way to avoid it.

I have previously ranted about this in some detail here:

http://lists.debian.org/debian-user/2007/08/msg00022.html

And more recently with a little more rationale here:

http://lists.debian.org/debian-user/2008/04/msg02554.html

From the SAR logs we can see that there is lots of memory in use by the
system cache, but free is low. How can we ensure there is more memory
available in free to avoid triggering oom-killer?

The easiest solution is to add swap space. That will increase your
virtual memory. Then you won't run out of memory and will avoid the
out of memory killer.

total used free shared buffers cached
Mem: 8115 7968 146 0 23 5575
-/+ buffers/cache: 2369 5745
Swap: 972 127 845

That is a very small amount of memory. I wouldn't have guessed that a
system could run with that small of memory these days.

Bob

Attachment: signature.asc
Description: Digital signature

Relevant Pages

  • Re: Improving OOM killer
    ... I'd like to suggest some changes to the OOM killer code that I believe ... which is to kill ... has failed to free any memory. ... parallel make in doc/ subdirectory of sources of a KDE module. ...
    (Linux-Kernel)
  • Re: Memory overcommit
    ... When preparing KSM unmerge to handle OOM, I looked at how the precedent ... of the same size as physical memory, ... by the poor decisions the OOM killer made. ... But this is then liable to hang the system if there's some reason ...
    (Linux-Kernel)
  • Re: Improving OOM killer
    ... set of allowed nodes with current, preempting the oom killer for GFP_DMA ... Besides the vague "forkbomb" description I still don't know ... spawning of children of a specific parent that'll lead to OOM. ... another task consuming much more memory could be chosen instead). ...
    (Linux-Kernel)
  • Re: [PATCH 1/2 (repost)] mm: serialize OOM kill operations
    ... __alloc_pagesthis will cause the task to retry the memory allocation ... by the OOM killer's chosen victim. ... easily with the modified 2.6.9-based redhat kernel that I've been mostly ... I've been tinkering with the OOM killer lately because we're moving ...
    (Linux-Kernel)
  • Re: Hyper-Threading Vulnerability
    ... The OS could do the scheduler tricks to avoid scheduling unfriendly ... to avoid *ever* referencing memory with the address determined by ... I can't think of any reason that this technique wouldn't work to ... If SHA has plaintext-dependent memory references, ...
    (Linux-Kernel)