> On Tue, 2009-04-07 at 13:11 +0100, Nick Lunt wrote:
> > Hi Tom
> >
> > yes it's running 64Bit ;)
> >
> > We have turned swap off for the time being...
> >
> > [r...@findb ~]# free -lom
> >              total       used       free     shared    buffers
> > cached
> > Mem:         64453      64147        306          0        123
> > 11782
> > Low:         64453      64147        306
> > High:            0          0          0
> > Swap:            0          0          0
> 
> Does removing swap actually eliminates the problem?  I'm a little
> surprised since kswapd really does more than just swapping, it
> basically
> scans memory whenever the number of free pages drops below a given
> threshold looking for memory to free.  Swapping is just one way it can
> free memory, it can also shrink the buffer and page cache, and find
> discardable pages.  That being said, it's possible that removing swap
> may change the dynamic of your systems memory management in others
ways
> that will keep it from happening.
> 
> I don't have any 64GB systems, but we have a couple of 32GB systems
> running Oracle and have never seen anything like this.  It does seem
> like your free pages are very low while your cache is over 11GB.  Does
> this have something to do with the 200,000 files you were talking
> about.
> Is whatever your doing with those causing continuous memory pressure
by
> dirtying the cache thus meaning that kswapd is constantly waking
trying
> to find memory to free?
> 
> Are you using huge pages with your Oracle setup?
> 

Hi Tom

I have just enabled a 2GB swap partition as a safety net. 
We have not tried running without swap before so I do not know if this
would fix the problem.

We are using huge pages 
        vm.nr_hugepages=15363

I aren't sure what the cached column report by free means to be honest,
could you give me a quick rundown ?

Here's how free looks with the 2GB swap device added:

[r...@findb ~]# free
             total       used       free     shared    buffers
cached
Mem:      66000400   65692088     308312          0     159324
15332996
-/+ buffers/cache:   50199768   15800632
Swap:      2097144        724    2096420

And here's a quick sar -r output

[r...@findb ~]# sar -r 3 5
Linux 2.6.18-92.el5 (findb.wales.nhs.uk)        04/07/2009

03:17:10 PM kbmemfree kbmemused  %memused kbbuffers  kbcached kbswpfree
kbswpused  %swpused  kbswpcad
03:17:13 PM    305180  65695220     99.54    157768  15277388   2096420
724      0.03         0
03:17:16 PM    306128  65694272     99.54    157644  15285480   2096420
724      0.03         0
03:17:19 PM    300444  65699956     99.54    157284  15284780   2096420
724      0.03         0
03:17:22 PM    306880  65693520     99.54    156100  15283400   2096420
724      0.03         0
03:17:25 PM    300152  65700248     99.55    155908  15285424   2096420
724      0.03         0
Average:       303757  65696643     99.54    156941  15283294   2096420
724      0.03         0

If any of this rings any bells please let me know :)

Thanks
Nick.

 
 

__________ Information from ESET NOD32 Antivirus, version of virus
signature database 3992 (20090407) __________

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com
 

_______________________________________________
rhelv5-list mailing list
rhelv5-list@redhat.com
https://www.redhat.com/mailman/listinfo/rhelv5-list

Reply via email to