[Linux-cluster] Tuning GFS and fnctl locks

[Peter Shearer]

Hi, All --
 
I'm playing around with FC4 (x86_64smp - Dell PowerEdge 1850 hardware)
and the GFS rpms included in the distribution.  Couple of questions:
 
I know that there's a list of tunable options given by gfs_tool gettune
<mountpoint>.  Is there a list somewhere of what these options do?
 
Is there a way to optimize GFS with regards to opening/closing files
with fnctl locks?  My app does a lot of writing with fnctl locks
(file-based database application), and it really drags on GFS when
running distributed locks.  I've posted about this in the past; I'm just
now getting around to looking at GFS again.  I haven't seen a problem
with regards to actual I/O throughput over my iSCSI SAN - my app under
GFS typically will do 30 kB/s, and I have been able to push 100+ MB/s
through the SAN - it's just slowing down significantly when the files
are opened with a fnctl lock.  The top command gives the following:
 
  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
 5996 root      18   0  4268 2700 1564 R 39.8  0.1   0:01.20 runcobol
 3201 root      10  -5     0    0    0 S 22.9  0.0   8:20.86 dlm_astd
 4285 root      10  -5     0    0    0 S 15.9  0.0   0:43.59 lock_dlm2
 4284 root      10  -5     0    0    0 S 13.9  0.0   0:44.91 lock_dlm1

I've seen dlm_astd and the two lock processes go higher than this in CPU
time, sometimes up to 30%+.
 
For some reason I remember reading somewhere that I should disable
hyperthreading when using GFS.  I can't seem to find where I read that
statement.  Am I'm just going insane, or is that really the
recommendation?
 
We'd like to see three or more of the machines in the mix; currently
there is only one.  Any suggestions would be greatly appreciated.
 
Thanks!
 
--Peter
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20050531/0637ec46/attachment.htm>