We used Gluster as our database filesystem for years. It was slow, and occassionally got into a split-brain state, but had very few serious problems. I would make sure that Gluster isn't reporting any errors, and also that you really only have one qmaster running (this was the cause of our only real serious problem - Gluster was in a split-brain, and someone got confused and tried to start a qmaster when one was already running).
Now we're on GPFS and are much happier. It's overkill for GE but we already have it available for our research and home directory filesystems. On Mon, Jul 09, 2018 at 03:48:20PM +0200, Paul Paul wrote: > Hello, > > In order to use the shadow master functionality, the SGE local configuration > files have to be stored on NFS. This is usually done by using a single > server, thus implies a single point of failure. > > If you're using SGE with a distributed file system, can you please indicate > which one? > We tried GlusterFS (version 3.12) with SGE 8.1.9 but it appeared that jobs > were randomly killed (after few days where everything run smoothly); going > back to NFS (4.0) on a single server fixed this behavior. > > Thanks for sharing, > > Paul. > _______________________________________________ > users mailing list > users@gridengine.org > https://gridengine.org/mailman/listinfo/users -- -- Skylar Thompson (skyl...@u.washington.edu) -- Genome Sciences Department, System Administrator -- Foege Building S046, (206)-685-7354 -- University of Washington School of Medicine _______________________________________________ users mailing list users@gridengine.org https://gridengine.org/mailman/listinfo/users
