[CentOS] GFS2 performance

Arnts, Leo

Leo.Arnts at PinkRoccade.nl
Fri Apr 22 07:01:04 UTC 2011


Hi,

I'm trying to get more performance out of my DRBD cluster with gfs2.
It seems that our gfs2 implementation is quit slow.

When running the ping_pong test we get no more than a 1000 locks/sec on the disk.
./ping_pong /mnt/backup/test.dat 4
879 locks/sec

The cluster config has been updated with:
        <dlm plock_ownership="1" plock_rate_limit="0"/>
        <gfs_controld plock_rate_limit="0"/>

gfs2_tool getargs /mnt/backup
statfs_percent 0
data 2
suiddir 0
quota 0
posix_acl 0
upgrade 0
debug 0
localflocks 0
localcaching 0
ignore_local_fs 0
spectator 0
hostdata jid=1:id=2752514:first=0
locktable
lockproto

gfs2_tool df -H
/mnt/backup:
  SB lock proto = "lock_dlm"
  SB lock table = "cybercenter:backup"
  SB ondisk format = 1801
  SB multihost format = 1900
  Block size = 4096
  Journals = 2
  Resource Groups = 7451
  Mounted lock proto = "lock_dlm"
  Mounted lock table = "cybercenter:backup"
  Mounted host data = "jid=1:id=2752514:first=0"
  Journal number = 1
  Lock module flags = 0
  Local flocks = FALSE
  Local caching = FALSE

  Type           Total          Used           Free           use%
  ------------------------------------------------------------------------
  data           3T             1T             1T             47%
  inodes         1T             96M            1T             0%

Hdparm information of the disk:
hdparm -tT /dev/sdb
/dev/sdb:
Timing cached reads:   20988 MB in  2.00 seconds = 10512.28 MB/sec
Timing buffered disk reads:   74 MB in  3.05 seconds =  24.30 MB/sec


Has someone an idea about increasing the performance of my gfs2 filesystem?

Kind Regards,
Leo Arnts


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.centos.org/pipermail/centos/attachments/20110422/c988ab21/attachment.html>


More information about the CentOS mailing list