[Bioclusters] GRIDEngine set up

chris dagdigian bioclusters@bioinformatics.org
Fri, 07 Feb 2003 11:19:34 -0500

andy law (RI) wrote:
> but if anyone out there is prepared to
> pitch in with their observations, I'll be very grateful.

Hi Andy,

Local caching of blast databases to avoid NFS is fine but removing NFS 
totally from a GridEngine setup may be overkill unless you have some 
other serious internal requirements/drivers for a no-filesharing 
deployment. A non-NFS setup will complicate your test cluster, 
especially in the early stages as you find yourself making constant 
changes to queue configurations and various SGE settings. It will also 
prohibit you from using the SGE failover features.

The path of least resistance in this case is to go ahead with NFS 
avoidance techniques for your data and blast work but leave a shared 
filesystem in place for the sole use of (a) grid engine and (b) 
(possibly) writing your blast results back to a common location.

I'd recommend doing this if at all possible. If you are in an 
environment where the shared filesystem is impossible then you can 
follow the NFSreduction howto that has already been posted. Let us on 
the list know how it goes!