TruCluster
cancel
Showing results for 
Search instead for 
Did you mean: 

vi problem opening file on NFS mounted files system

vi problem opening file on NFS mounted files system

We have two clusters A and B. Each of these clusters has two systems. Both clusters are running Tru64 version 5.1A patch kit 2. Cluster A does a NFS mount of a file system on cluster B. After a re-boot of cluster B the cluster_lockd daemon has a state of ONLINE:

# caa_stat -t
Name Type Target State Host
------------------------------------------------------------
cluster_lockd application ONLINE ONLINE B1

However, when I try to use the vi editor from cluster A to access a file on cluster B over the NFS mounted file system, the response time to open the file is extremely long. Some users think vi is hanging. A work-around I found is, that when I stop, start, or relocate the cluster_lockd on cluster B until it is in a state of UNKNOWN the response time problem is solved. After this ???juggling??? opening a file with vi happens immediately.

# caa_stat -t
Name Type Target State Host
------------------------------------------------------------
cluster_lockd application ONLINE UNKNOWN B1

I did not find any reference to this state. What is this state of UNKNOWN? Is there another way to address the vi response time problem?
7 REPLIES
Dave Bechtold
Respected Contributor

Re: vi problem opening file on NFS mounted files system

Hi,

The UNKNOWN state under CAA means that the "stop" entry point in the action script failed or exceeded it's time-out. So, from CAA's perspective it doesn't know if cluster_lockd actually started or not, hence UNKNOWN. To get it back to ONLINE you must first stop the application and then use caa_start to re-start it again.

Does this problem happen for all vi sessions all the time, or intermittently ?

Or does it happen for only the first vi session ?

How is NFS setup on the cluster?
- Are you using autofs or auotmount?
I'd advise to use autofs for v5.1 and higher.

- Are you using cdsl mountpoints for the nfs mounts?

If not I'd recommend that, checkout the Best Practice -> http://h30097.www3.hp.com/docs/best_practices/BP_NFSCDSL/TITLE.HTM .

Hope this helps,
Dave Bechtold



Ralf Puchner
Honored Contributor

Re: vi problem opening file on NFS mounted files system

please check if a rpc.lockd and rpc.statd is running on each member. The member with the caa lock service requires an additional rpc.lockd and rpc.statd.
Help() { FirstReadManual(urgently); Go_to_it;; }

Re: vi problem opening file on NFS mounted files system

Thanks for the replies.

- Does this problem happen for all vi sessions all the time, or intermittently ?

Yes, it happens for all vi sessions.

- Are you using autofs or auotmount?
No, we are not using autofs. Will set it up and see if it helps.

- Are you using cdsl mountpoints for the nfs mounts?
No, we are not using cdsl. Will do this too. Thanks for the best practice reference.

- Please check if a rpc.lockd and rpc.statd is running on each member. The member with the caa lock service requires an additional rpc.lockd and rpc.statd.
Yes, These are running:
# ps -ef | grep rpc*
root 1049512 1048577 0.0 Aug 21 ?? 0:00.00 /usr/sbin/rpc.statd
root 1049515 1048577 0.0 Aug 21 ?? 0:00.00 /usr/sbin/rpc.lockd
root 1060824 1048577 0.0 Aug 22 ?? 0:00.00 /usr/sbin/rpc.statd -c
root 1060827 1048577 0.0 Aug 22 ?? 0:01.95 /usr/sbin/rpc.lockd -c

Mathias
Ralf Puchner
Honored Contributor

Re: vi problem opening file on NFS mounted files system

Have you checked the settings on both clusters? Locking requires an lockd/statd on both sides.

Any error messages within the error logs?

What happened if you create a new file via vi?
Does it hang or only on existings files?
Help() { FirstReadManual(urgently); Go_to_it;; }

Re: vi problem opening file on NFS mounted files system

To clarify, the system is a GS160 ??? GS80 production cluster which mounts a file system from a ES40 ??? DS20 cluster. From a user perspective everything works fine. NFS read and writes are fine. There are no user problems reported
Nevertheless, the cluster_lockd on the ES40 ???DS20 cluster reports this ???UNKNOWN??? state and I was simply curious what it means.
Yes, after a re-boot, the cluster_lockd on the ES40 ???DS20 cluster reports a ???ONLINE??? state and then vi session from the GS160 ??? GS80 cluster to the ES40 ??? DS20 cluster are hanging. I don???t remember if vi also hangs with new files. I have no means to test it because it can take weeks until the production cluster will be re-booted. But again, thanks for the responses.
Ralf Puchner
Honored Contributor

Re: vi problem opening file on NFS mounted files system

nfs and locking are different things and mechanism. So if nfs works this gives us no clue if locking is still working.

It seems your locking doesn't work properly. So please verify if

server:
a) one rpc.lockd and rpc.lockd per member is running
b) one additional rpc.lockd -c and rpc.statd -c is running at the owner of the caa lockd resource

client:
rpc.lockd is running
rpc.statd is running

If you can create a new file on the server and doesn't have any problem but opening an existing file fails, problem relies to locking.

So please check the processes and have a look into /var/adm/syslog.dated/current/daemon.log if problems will be logged on the client or server.
Help() { FirstReadManual(urgently); Go_to_it;; }
Agnes Coat
Advisor

Re: vi problem opening file on NFS mounted files system

Check the 2 CDSLs
/etc/sm
/etc/sm.bak

Agnes