NFS and Tru64 5.1

[Michael Greulich]

Hi folks,

we have a problem mounting a NFS directory to a Tru64 5.1 client. The NFS server is running on Solaris 8 and serving other clients (including Tru64 4.0) very well. Just 2 boxes on Tru64 5.1 2650 have the problem that files cannot be edited. When you try starting vi with a file on the NFS share vi just locks up. Creating new files work for some reason?!?

Any suggestions?

TIA
Michael

[Johan Brusche]

Are these 2 V5.1B systems forming a cluster?

Are the rpc.statd and rpc.lockd running ?
If in TruCluster is CAA service cluster_lockd on-line ?

What's in /etc/exports on the Solaris system ?

___ Johan.

[Michael Greulich]

Yes, these two systems are V5.1B and running TruCluster. rpc.statd and rpc.lockd are running. I don't know how to check the cluster service because i'm not so familiar with TruCluster. At least there war is no process running and no entry with this name in /etc/services or /etc/inetd.conf

This is the exports line from Solaris
share -F nfs -d "Homedirectories" /d/home
and this is the corresponing line on Tru64
tbmd02:/d/home /d/home nfs rw,bg,hard,intr 0 0

All the options rw,bg,hard,intr should be ok, because they are set on the V4.0 Clients as well.

.\\ichael

[AwadheshPandey]

do
#sysman nfs
and view status of nfs services on cluster, and restart the service if there is any error post the result.

you can configure ur nfs by this way,

Awadhesh

[Al Licause]

A ps command will show the member specific as well as the cluster wide statd/lockd daemons:

# ps ax | grep rpc
1056226 ?? I 0:00.02 /usr/sbin/rpc.statd
1056228 ?? I 0:00.04 /usr/sbin/rpc.lockd
1109653 ?? I 0:00.03 /usr/sbin/rpc.lockd -c
1109655 ?? I 0:00.03 /usr/sbin/rpc.statd -c

Those processes with the -c are the cluster wide lock daemons.

You might want to use tcpdump on the tru64 unix host or snoop on the Sun system to trace this connection and see who is waiting on who.

[Aco Blazeski]

Hi Michael,

be sure that in exports on Solaris machine you specify the cluster name/IP. That is:

tbmd02:/d/home /d/home nfs rw,bg,hard,intr 0 0

tbmd02 should be the clustername and it should be defined in /etc/hosts on Solaris machine. This because each tru64 cluster member represents itself in the network with cluster IP.

You can obtain this info issuing these commands on some tru64 cluster member:

[code]
tom:root# clu_get_info -full
Cluster information for cluster mgm
...
tom:root# grep mgm /etc/hosts
IP.add.of.cluster mgm
tom:root#

[/code]

cluster name is "mgm" and its IP is "IP.add.of.cluster"

also , to check cluster_lockd

[code]
tom:root# caa_stat -t
Name Type Target State Host
------------------------------------------------------------
...
cluster_lockd application ONLINE ONLINE tom

[/code]


Hope this will help

[Michael Greulich]

Thanks for all of your answers.

I've checked all the processes and the cluster appears to be up and running. The problem might be that the NFS mounts should be active on both members doesn't matter whether the cluster application is running on the node or not (homedirs). So i guess that it locks up at some stage where a mixup with cluster and real IP happens. I will doublecheck this with another colleague today.