[Linux-cluster] GFS file size/performance question

Jeff Sturm jeff.sturm at eprize.com
Wed Apr 15 14:47:40 UTC 2009


> -----Original Message-----
> From: linux-cluster-bounces at redhat.com 
> [mailto:linux-cluster-bounces at redhat.com] On Behalf Of David Teigland
> Sent: Tuesday, April 14, 2009 11:47 AM
> To: linux clustering
> Subject: Re: [Linux-cluster] GFS file size/performance question
> 
> On Mon, Apr 13, 2009 at 10:38:15PM -0400, Jeff Sturm wrote:
> 
> > gfs_tool settune $fs glock_purge 50
> 
> > There have been many prior articles posted on GFS 
> performance.  Search 
> > for e.g. "glock trimming patch".
> 
> That purge/trimming thing was a hack for a very particular 
> corner case (backup process walking over the entire fs and 
> caching all locks on the fs).  Enabling it under other 
> circumstances will most likely *hurt* your performance by 
> artificially reducing normal gfs caching.

In our case it was a periodic rsync process.  The OP was asking about
large numbers of files, so he may have had something similar.  The
gradual slowdown of a single node sounds a lot like what we had seen
with rsync.

Jeff





More information about the Linux-cluster mailing list