[Linux-cluster] Clusterng with GFS

[Scott McClanahan]

On Wed, 2007-05-30 at 08:27 -0400, Kevin Anderson wrote:
> On Wed, 2007-05-30 at 09:29 -0400, Daniel Fernanduz wrote: 
> > Hello
> > 
> > I am trying to implement a failover cluster with shared storage concept 
> > (Postgresql Database 7.4).
> > RHEL AS 4.0 + RHCS + GFS
> > 
> > I am having two gfs volumes which gets mounted by cluster when any of 
> > the nodes in
> > cluster takes control. The volumes are getting mounted by the cluster 
> > whenever a node takes
> > control, but not getting unmounted when the node leaves the cluster. 
> > Especially when primary
> > node takes control, the volumes mounted on the secondary node (Node 
> > which gives control to the
> > primary node) is not getting unmounted. Anybody plz help me to solve 
> > this issue. Thanks for the help in advance
> 
> The default filesystem failover script special cases gfs filesystem
> types.  It knows that gfs is a cluster filesystem and you don't need
> to mount/umount it in a failover environment.  Typically failover
> environments that utilize gfs are doing it to improve failover times
> by not having to mount/umount the filesystem.  That is one of the
> benefits of having a cluster filesystem.  If you want to just do
> failover the (old)slow way, then either use ext3 or modify the
> failover script.
> 
> Kevin
> 
> plain text document attachment (ATT268870.txt), "ATT268870.txt"
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster

It's nice having the file system mounted on multiple nodes but isn't GFS
I/O a bit slower since file system locks have to be requested from the
first node that mounted the file system?