bdf still hangs after package halt despite change to UDP

[Ian Vaughan]

Howdy,
When I stop my SAP package that has the HANFS mounts in it, bdf is seen to hang with "NFS Server not responding still trying".

In accordance with the tech doc
"HP-UX Serviceguard/SAP: long SAP-package switch time"
I have changed the protocol that the automounter (autofs) uses to UDP from TCP in the mount options in /etc/auto_master. I have done an nfsstat -m to make sure that UDP is the protocol used, which it is.
I am now stuck as to what to do to make bdf respond in a timely manner once the package has halted.

Any comments or suggestions are welcome.
Regards
Ian

[Michael Steele_2]

Hi Ian:

a) Can you stop and restart the NFS server or clients?

b) Are you using current file system patches, i.e., online JFS?

c) Do you see the same problem with 'fstyp' and 'df'?

d) You've checked for disk and file system errors in syslog.log?

[Ian Vaughan]

Hello,

a) Can you stop and restart the NFS server or clients? Yes - and nfs.core, still hangs

b) Are you using current file system patches, i.e., online JFS? - 11iV2 patched up to Sept 2006

c) Do you see the same problem with 'fstyp' and 'df'? df- yes, same as bdf. fstyp gives me "no such device or address" on the lvol presumably because the VG is offline.

d) You've checked for disk and file system errors in syslog.log? Yes - it seems purely to be with the fact that the directories that auto.direct had mounted are still in /etc/mnttab after the package has been halted and the disks have gone.

I have changed to UDP proto for the mounts but the dirs are still in mnttab and bdf hangs just the same.

Thanks for trying tho

Regards
Ian

[Raj D.]

Lan,

As the bdf is hanging so its sure it is nfs related problem that happend while package halt.


Can you check the package log file pkg.cntl.log for any error. And if any difference is there in the log file , other than usual way of halting the package, like unable to unmount some nfs mount points or so. Then it might help in debuging the bdf hanging problem.


Also you can check :

# bdf -l
# strings /etc/mnttab
Any difference ??

Cheers,
Raj.

[rick jones]

Not that it will (directly) solve anything, but you might try taking a tusc trace of the bdf command to see in which system call it is hanging.

[Ian Vaughan]

Howdy all,
bdf -l works fine as presumably it only looks at local filesystems.

/etc/mnttab still has the automounted directories in it.

nfsstat -m still shows the mounts

I am now thinking that this is a characteristic rather than a fault. I could reduce the timeout on the automount side so that the dir is unmounted after a shorter time period without access.

Cheers
Ian