[ClusterLabs] Best way to failover NFS
dennisml at conversis.de
Tue Jun 9 18:29:49 EDT 2015
I'm wondering what the best way is to failover a NFS system. As far as I
understand there are two ways to do this:
1. Put /var/lib/nfs on a redundant device and start the nfs server only
on one node. Exports can be statically defined on both systems and will
become automatically active the moment nfsd is started.
2. Use exportfs resources and keep nfsd running on both nodes. When a
failover occurs the exportsfs resources are used to dynamically
re-export the the endpoints on the new system.
What is the best method? 1 seems cleaner to me because there is no need
to manage a bunch of exportfs resources in Pacemaker and the metadata
from the failed node is available on the new one.
Another question is how to deal with the nfsv4 lease system. When using
the exportfs method I know that you have to wait $gracetime seconds
before unexporting a filesystem so the clients get a new lease from the
new system but is this also required when using method 1 since there the
metadata is mirrored on the new system?
More information about the Users