Showing results for 
Search instead for 
Did you mean: 

How to get a different member to handle NFS

Victor Semaska
Frequent Advisor

How to get a different member to handle NFS


We have a 5-member Tru64 cluster, V5.1b PK3. We're having a problem on one of the members, in this case member 2, and would like to move services off of it one at a time to help determine what the problem is. The problem is an application (Banner/Oracle) is occasionally experiencing timeouts.

The cluster is set up as a NFS server and we've noticed that NFS requests are going to this member. How do we relocate it so those NFS requests are handled by a different member.

CAA resoucre cluster_lockd was on member 2 so I used caa_relocate to put it on another member. NFS requests are still being handled by member 2 (according to our network people who are using sniffers to monitor network traffic on that member).

arp -a shows the cluster alias still on member 2. I assume that's why NFS requests are still handled by it. I looked at the manpage for cluamgr and found the following command:

# cluamgr -a alias=alias,leave

If I use this command on member 2 would it cause the NFS requests be handled by another member? I've never used this command before and this is a production cluster, that's why I'm reluctant to try it.

Ralf Puchner
Honored Contributor

Re: How to get a different member to handle NFS

The answer is within the cluster admin guide. Please go to the chapter describing how NFS works and aliase are managed on clustered members and you will be able to answer the question yourself.

Btw. there are 2 different ways NFS works (UDP/TCP) leading to totally different NFS request handlings. So what kind of connections are used by your NFS clients?

PS: there is also a config file defining what member/alias handles NFS requests...

Help() { FirstReadManual(urgently); Go_to_it;; }
Johan Brusche
Honored Contributor

Re: How to get a different member to handle NFS

If the alias you are talking about is the default alias, other services on that member will stop to work too. If it's a secondary alias only used by NFS-clients for mounting, then that command is OK.

Or... why no use '/sbin/init.d/nfs stop' and if you want to make that more permanent:
'rcmgr -h x set NFSSERVING 0' (where x=memberid)