NFS automount timeout issues

[Craig Irvine]

Hi All,

We are setting up 2 systems for production SAP in a ServiceGuard environment. The systems are HP-UX 11.23 and so have the HA AutoFS updates. /etc/rc.config.d/nfsconf on both nodes (node1, node2) have:
AUTOMOUNT_OPTIONS="-t 10"
AUTOMOUNTD_OPTIONS="-L"

I have set up a simple test package (tstpkg)that has a single logical volume that is setup to automount to both nodes. The automounting is working fine until I halt the package on node1 and start it on node2.

I run the following commands (prefixed by the node name):
node1: cmrunpkg -n node1 tstpkg
node1: cd /mnt/tstdir1 #works instantly
node1: cd /
node2: cd /mnt/tstdir1 #works instantly
node2: cd /
node2: cmhaltpkg tstpkg
node2: cmrunpkg -n node2 tstpkg
node2: cd /mnt/tstdir1 #approx 10 sec. delay
node2: cd /
node1: cd /mnt/tstdir1 #hangs for 10 min. then cd's OK

syslog.log on node1 has the following message from that last cd:
Jan 11 16:19:06 node1 vmunix: NFS server tstpkg not responding still trying
Jan 11 16:28:13 node1 vmunix: NFS server tstpkg ok

The first message appears 1 minute after issuing the last "cd /mnt/tstdir1" command on node1.

Ultimately the whole thing works but the 10 minute delay shouldn't be happening. Does anyone know the reason for the delay??

Thanks for any help.

Regards
Craig

[Peter Godron]

Craig,
Take a look at ITRC docs NETUXKBRC00006283 and KBAN00000261.
This may give you a few new pointers.
Regards

[Craig Irvine]

Thanks Peter, they are a couple of very good documents and provided hints at other documents to look at.

The problem wasn't performance related as I am the only person/process using the 2 nodes (let alone using NFS on the nodes) and I was just cd'ing to the automounted directories. The issue seemed to be with stale entries in /etc/rmtab which mountd uses to populate its tables when it starts.

For others that may have a similar problem I followed the following steps:
- shutdown package(s) using NFS
- /sbin/init.d/nfs.client stop
- /sbin/init.d/nfs.server stop
- /sbin/init.d/nfs.core stop
- /sbin/init.d/Rpcd stop
- check for any stale nfs related processes
- clean out /etc/rmtab
- /sbin/init.d/Rpcd start
- /sbin/init.d/nfs.core start
- /sbin/init.d/nfs.server start
- /sbin/init.d/nfs.client start
- start packages and test if necessary

I can now swing the package between the nodes at will and the automount connects OK.

Regards
Craig

[Craig Irvine]

As per my last post.