rpc.lockd took excessive CPU usage

[Yan Berube]

Hi,

I have a L3000 with a lot of user connect to it. And I saw that the process of rpc.lockd took 90 % of my CPU. Can I kill this process ? If i do this, it have an effect on the user workstation ? How I can fix this issues without affecting the production environment ? It's hp-ux 11.00 install on this server.

Thanks for you help, it's very appreciated !

[Sridhar Bhaskarla]

Hi,

rpc.lockd is required for NFS activity. It takes the requests to lock the files from the remote server and sends lock requests to the remote servers.

You shouldn't kill it if you are using NFS. If you are not using NFS, then disable NFS using /sbin/init.d/nfs.* scripts.

If you use NFS, then try to install latest NFS/ONC General Release performance patch.

-Sri

[Brian Hackley]

Yan,

rpc.lockd is probably being used for file locking of NFS mounted home directory shell history files.

Workaround for user:
put shell history files in local directory adding to /etc/profile or .profile:
export HISTFILE=/tmp/.sh_history_$LOGNAME

We suggest you check that the remote hostnames in /var/statmon/sm and /var/statmon/sm.bak can be looked up with the command:
nsquery hosts
This will use your nsswitch.conf file to look up the remote hostnames.
Then ping each hostname!!
Remove any entries that are suspect.

Also you can put rpc.lockd into debug mode and review the output. Get the PID of the running rpc.lockd, and sent it a SIGUSR2 signal to turn on debug, sent another one to turn it off...e.g. kill -17 pid
The debug output is in /var/adm/rpc.lockd.log by default. Careful, this will grow quickly. Only do this if all other steps fail since you may need HP Support to help interpret the output.

Latest ONC Cumulative patch is PHNE_29210 + dependencies!

Hope that helps,
-> Brian Hackley