[rhelv6-list] RHEL6.2 XFS brutal performence with lots of files
Pat Riehecky
riehecky at fnal.gov
Mon Apr 15 14:09:09 UTC 2013
On 04/15/2013 09:02 AM, Daryl Herzmann wrote:
> Thanks for your help.
>
> On Mon, Apr 15, 2013 at 8:39 AM, Pat Riehecky <riehecky at fnal.gov
> <mailto:riehecky at fnal.gov>> wrote:
>
> I've run into some terrible performance when I've had a lot of
> add/remove actions on the filesystem in parallel. They were mostly due
> to fragmentation. Alas, XFS can get some horrid fragmentation.
>
> xfs_db -c frag -r /dev/<node>
>
> should give you the stats on its fragmentation.
>
>
> # xfs_db -c frag -r /dev/md127
> actual 140539575, ideal 139998847, fragmentation factor 0.38%
>
> Here's an iostat snapshot while I was running xfs_db, the tps numbers for
> md127 seem strange. sd[b-f] are a part of the raid5....
>
> avg-cpu: %user %nice %system %iowait %steal %idle
> 5.12 0.00 7.12 8.81 0.00 78.94
>
> Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn
> sda 35.00 1032.00 656.00 1032 656
> sdc 458.00 32451.00 680.00 32451 680
> sde 451.00 30936.00 544.00 30936 544
> sdb 573.00 32448.00 784.00 32448 784
> sdd 527.00 30728.00 696.00 30728 696
> sdf 593.00 31736.00 624.00 31736 624
> md127 157986.00 157983.00 1592.00 157983 1592
>
> I can't speak for others, but I've got 'xfs_fsr' linked into
> /etc/cron.weekly/ on my personal systems with large XFS filesystems.
>
>
> Seems like I shouldn't have to do that given the numbers above?
>
> daryl
>
I would agree, your drive looks nice and optimized..... this thread already
has you checking for lazy counts.... those are my two "tricks" for getting XFS
to run beautifully. But I've never had a volume over 8TB before.....
Pat
--
Pat Riehecky
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/rhelv6-list/attachments/20130415/8dd9e6f2/attachment.htm>
More information about the rhelv6-list
mailing list