[Linux-cluster] GFS vs Ext3/4

[jeremymiller at ups.com]

I want to get some opinions from the group.

In one of our development environments we have 80+ databases split across two cluster nodes. (yes it's a lot) Each database instance has 3-4 ext3 filesystems mounted on the node running the DB.  The databases are split across the cluster with roughly 40 DBs per node.  Having to maintain all these resources in the cluster.conf file is tedious and the file is _enourmous_.  In fact, we believe that we have seen this impact rgmanager on a number of occasions.  This is the primary reason for why we are considering GFS2 as opposed to ext3 - in that it _greatly_ reduces the clutter from the cluster.conf file and should alleviate the load on rgmanager.  However, and this is the reason for this email, is it fundamentally a mis-use of GFS2 to be using it when there is no requirement for a shared filesystem across the cluster nodes?  Or, is using GFS2, regardless of the requirement for share-access, the direction intended by the developers for all cluster services?  Granted, there will be a slight performance hit using GFS2 vs ext3 due to the locking overhead.  What other pros/cons are there to GFS2 vs ext3/4 when there is no real need for shared filesystem access across the cluster?

Regards,
--
JM