rpc.lockd -c coredumps (cluster_lockd)

[Gerard F Kilburn]

Hi

I have a 2 node cluster, TRU64, 5.1 pk3 which is nfs serving a few terabytes successfully. I have a problem though with the cluster_lockd continually failing, retrying, relocating to the second node and so on.

caa_stat -t shows cluster_lockd "ONLINE ONLINE"
and a ps shows
/usr/sbin/rpc.statd
/usr/sbin/rpc.lockd
/usr/sbin/rpc.statd -c

but there is NO "/usr/sbin/rpc.lockd -c"

I get a core file for rpc.lockd produced every time it tries to start with the -c flag. I've tried caa_stop followed by a caa_start and tried it on both nodes but whichever node it runs on, it core dumps.

What does rpc.lockd initiate when it starts in cluster mode? are there perhaps any files left hanging around that need housekeeping following a crash (the system had a power outtage previously), (I have tried replacing rpc.lockd with a 'clean' version from the same patch kit level without success) - note the rpc.lockd runs happily without the "-c" flag.

Thanks in advance for any assistance

Regards

Gerard.

[Johan Brusche]

I have recently seen the cluster_lockd service fail because the /etc/hosts file had stray non-printable chars (eg. ESC sequence).

Save your /etc/hosts, then grep out all lines starting with comment and all empty lines, then caa_stop cluster_lockd then start it.

Rgds,

___ Johan ./

[Gerard F Kilburn]

Thankyou for the reply Johan, I cleaned up my /etc/hosts file as you suggested but the problem is still there. (I also did a hex dump of it to look for odd characters but sadly found none).

Regards

Gerard

[Al Licause]

You're running a fairly old version of the OS and patch level.

I'd suggest at least applying the last patch kit for v5.1, but idealy upgrading to the latest OS version (v5.1b) with pk5.

Almost ever patch kit has some nfs and nfs related component patches.

Also try looking at the release notes for the various patch kits to see if something is mentioned that might address this.