[Linux-cluster] High DLM CPU usage - low GFS/iSCSI performance

Alan Brown ajb2 at mssl.ucl.ac.uk
Fri Feb 25 10:03:07 UTC 2011


On 25/02/11 07:21, Martijn Storck wrote:
> Thanks for your message.
>
> Somehow the issue has not returned since yesterday when I applied some 
> tuning to our GFS, specifically:
>
> glock_purge 50
> demote_secs 100
> scand_secs 5
> statfs_fast 1
>

It's most likely the biggest contributor is the statfs_fast setting

For GFS1, We had the following settings (which gave pretty good performance)

for d in `mount | grep ^/dev | grep gfs | grep -v gfs2 | cut -f3 -d" "`; \
  do \
  /sbin/gfs_tool settune $d glock_purge 80 & \
  /sbin/gfs_tool settune $d demote_secs 25 & \
  /sbin/gfs_tool settune $d statfs_fast 1 & \
  /sbin/gfs_tool settune $d statfs_slots 512 & \
  /sbin/gfs_tool settune $d quota_account 1 & \
  /sbin/gfs_tool settune $d quota_enforce 0 & \
  /sbin/gfs_tool setflag inherit_jdata $d & \
  done








More information about the Linux-cluster mailing list