NFS4 recovery capabilities in failover clusters

NFS4 recovery capabilities in failover clusters

[Matthias Eble]

Hi all,

I've been trying to find out how I can build an HA (active/passive)
NFS4 failover
cluster with smooth failure recovery. After a takeover, or when
clients return from isolation,
D-state processes and stale handles, etc should be prevented.
What I found were some linux-nfs wiki pages [1] about problems with recovery in
the linux implementation, and some hints about nfs 4.1's transparent
state migration.

What I couldn't find out is, how to actually use TSM, what it can do,
and what it can't.

So I have two questions:
* Have the issues mentioned in [1] been fixed and is the linux NFS4
client/server
  implementation currently able to provide solid recovery in a failover cluster?
  Just mount FS on other node + virtual IP?
* If so, do you have any hints, keywords, links etc with up-to-date information,
  what needs to be done to make such a cluster/recovery as resilient
as possible?
  Is just syncing v4recovery enough?

I've tried to get some definitive information on the net, the results
were either dated
or incomplete. And I guess the RFC does not necessarily apply to the
linux implementation, right?

Any help is appreciated.
Thanks for your time.
Matthias

[1] http://linux-nfs.org/wiki/index.php/NFS_Recovery_and_Client_Migration