RE: Linux LowMem errors

  • From: "David Hill" <David.Hill@xxxxxxxxxxxx>
  • To: <niall.litchfield@xxxxxxxxx>, <tim@xxxxxxxxx>
  • Date: Mon, 11 Feb 2008 11:41:47 -0500

I had a similar problem with OOM errors on a RH 3 box running an smp
kernel, with 9i RAC DB's

 

Every week I had DB's crashing with

ORA-27504: IPC error creating OSD context, and other OOM errors

 

Setting the following kernel parameter helped for while, then every few
weeks my load would shoot through the roof 150+, then die

vm.vm-defragment = 1000

 

Using the hugemem kernel has fixed the problem here, for now at least.

I'm migrating to 64 now.

 

and since your already on hugemem, any patches will only be bandaid
solutions.

The only way you truly fix the problem though is to migrate to 64bit

 

David Hill

Oracle DBA

 

 

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Niall Litchfield
Sent: Monday, February 11, 2008 5:20 AM
To: tim@xxxxxxxxx
Cc: Oracle List
Subject: Re: Linux LowMem errors

 

Not that MTU! There's an Oracle patch which has the following effect 

        change skgxp MTU from 32k to 16k

What seems to be happening is that Oracle requests 32k chunks of memory
from the OS for (I think) IPC packets. The bug description suggests that
Linux is somewhat prone to low memory fragmentation and isn't able to
allocate the memory. 

 

On Feb 10, 2008 3:59 PM, Tim Gorman <tim@xxxxxxxxx> wrote:

        MTU of 32768 sounds interesting -- usually it is set to 1500,
occasionally 9000.  Not sure if this is significant, and I could be
misinterpreting "requirement" for "setting" or some such, but I thought
worth noting... 

        
        
        
        
        Niall Litchfield wrote: 

        We've recently updated some of our RH database servers to kernel
2.4.21-53ELHugemem and have experienced problems with the databases and
ASM instances crashing and huge numbers (like 80,000) yesterday on one
box of core dumps. The crashes did not occur immediately after the
kernel update, but that is AFAIK the only significant recent change. The
head of the trace reads

         

        SKGXPVFYMMTU: Could not create context which satisfies the
minimum 

        MTU requirement of 32768 bytes

        kgefec: fatal error 0

        *** 2008-02-09 07:29:39.767

        ksedmp: internal or fatal error

        ORA-00603: ORACLE server session terminated by fatal error

        ORA-27504: IPC error creating OSD context

        ORA-27300: OS system dependent operation:sendmsg failed with
status: 12

        ORA-27301: OS failure message: Cannot allocate memory

        ORA-27302: failure occurred at: sskgxpsnd1

        
        which looks significantly like the issue addressed in Note
419871.1. to me. Has anyone else come across these issues? 
        -- 
        Niall Litchfield
        Oracle DBA
        http://www.orawin.info <http://www.orawin.info/>  




-- 
Niall Litchfield
Oracle DBA
http://www.orawin.info 


__________________________________
Nous avons déménagé! Notre nouvelle adresse est le :
We've moved! Our new address is:

8300 Décarie Boulevard
Montréal, (Québec)  H4P 2P5

CONFIDENTIALITY NOTICE
This message contains confidential information intended only for the use of
the individual or entity named as recipient. Any dissemination, distribution
or copying of this communication by anyone other than the intended recipient
is strictly prohibited. If you have received this message in error, please
immediately notify us and delete your copy. Thank you.

AVIS DE CONFIDENTIALITÉ
Les informations contenues aux présentes sont de nature privilégiée et
confidentielle. Elles ne peuvent être utilisées que par la personne ou
l'entité dont le nom paraît comme destinataire. Si le lecteur du présent
message n'est pas le destinataire prévu, il est par les présentes prié de
noter qu'il est strictement interdit de divulguer, de distribuer ou de
copier ce message. Si ce message vous a été transmis par mégarde, veuillez
nous en aviser immédiatement et supprimer votre copie. Merci.

Other related posts: