Editor "vi" hangs on NFS mounted directories

[Ramesh.K.R.]

Hi,

When we try to edit the files under NFS mounted directories, "vi" hangs permanently. What could be the problem ?? How to solve this problem ??

Thanks & Regards,
Ramesh.K.R.

[Agnes Coat]

Hi,

OS ? PK ? standalone or cluster ?

btw, check the directories /etc/sm and /etc/sm.bak (CDSL in V5.x)

Agnes

[Ralf Puchner]

this question was asked in the beginning of this forum. Please have a look into the old solution.

vi problem will be forced by NFS locking issues...

[Ramesh.K.R.]

Hi Agnes,

Sorry, i did not give full info.
We are using Tru64 5.1B, standalone type
What do you mean by PK ??

Regards,
Ramesh.K.R.

[Ramesh.K.R.]

Hi Ralf,

Thanks for reminding. But, when i serched the old archives, i was not able to get the solution/problem ?? Can you tell me more specifically ??
What do you mean by NFS locking ??

Regards,
Ramesh.K.R.

[Ralf Puchner]

Ramesh,

it means vi requires since 5.x of Tru64 a working nfs locking system (rpc.lockd) on server and client. So in case vi hangs the locking system doesn't work on your machine.
Use rpcinfo -p to check if daemons are running on both sides.

pk means patchkit. It is important you have installed the current patchkit fixing some locking issues.

[Ramesh.K.R.]

Hi,

We have the patch kit T64V51BB22AS0002-20030415 installed. Do you mean to say, nfs.lockd should be running on the system as a daemon ??

Regards,
Ramesh.K.R.

[Ramesh.K.R.]

Sorry,

what i ment was "rpc.lockd"

Ramesh

[Ralf Puchner]

have you tried the given command "rpcinfo"?
it should list the rpc.lockd process (also a ps should list the process).

If you don't know what nfs locking is, please consult the man-page of rpc.lockd.

[Johan Brusche]

Let's apply some basic NFS troubleshooting:

1) What is the OS and version and patchkit (aka PK) level of the system
exporting the filesystem via NFS.

2) What is the OS and version and patchkitlevel of the client
that NFS mounts this filesystem.

3) If the OS is Tru64 UNIX, what is the result of
rcmgr get NFSLOCKING
rcmgr get NUM_PROPLISTD

3b) If the result of above commands is non-zero, does the system still run rpc.lockd,
rpc.statd and proplistd
(proplist is only required if exported filesystem is using filequota and or ACL's)

3c) If this is on a TruCluster V5.x, is one of the members running "rpc.lockd -c" and
"rpc.statd -c", ie on the member that shows
cluster_lockd ONLINE with caa_stat

4) As already said in above replies use:

rpcinfo -p | grep lock
rpcinfo -p | grep lock

to check if nfslocking is registered via the portmapper (is process portmap running ?)

5) As mentionned in the first reply, the existence of the directories /etc/sm and /etc/sm.bak is required

Please always provide as much as possible of above info when posting NFS related problems.

Rgds,
Johan.

[ROCH_2]

I've got the same problem with vi handling on NFS FS.

My configuration :

1)T64V51AB21AS0004-20030206 OSF520

2)T64V51AB21AS0004-20030206 OSF520

3)rcmgr get NFSLOCKING =1
rcmgr get NUM_PROPLISTD =0

on both servers

3b)my system still run rpc.lockd, rpc.statd (no filequota or ACL)

3c)No Cluster

4) nfslocking is registered via portmapper

5) /etc/sm and /etc/sm.bak exists


I can give you a detail with my investigations : This should another server (that is not using the NFS server) that cause this problem on the machine. I'm trying to reproduce the problem.

If somebody have an idea!

[Ralf Puchner]

is it possible, that you or your server have more than one network interface or address?

[Mathias Schmassmann]

I asked this question in the beginning of this forum, however, there was never a real solution posted.

My work-around to make vi work on NFS mounted file systems was to relocate the cluster_locked daemon between members of the cluster until it is in a state of UNKNOWN (example with a cluster with two systems A1 and B1):

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

# caa_relocate cluster_lockd â c A1
# caa_relocate cluster_lockd â c B1
# caa_relocate cluster_lockd â c A1
# caa_relocate cluster_lockd â c B1


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

[ROCH_2]

My server as only One NIC and One adress.

[Ralf Puchner]

have you checked name resolution (both sides must see the same name) and have you verified proper nfs locking?

[ADJOLOHOUN]

Hi,

Check daemon.log for error :

"lockd[pid]: Can't create client handle to nfs-server NLMv4: RPC: Program not registered"

it means that rpc.lockd can't bind to nfs-server.

Please kill rpc.lockd, and try :
# /usr/sbin/rpc.lockd -b NFS-SERVER &

Fx Adjolohoun.

[ROCH_2]

Hi,

So I think there's no problem with name resolution. Everything is OK in /etc/hosts.

I've got errors in the daemon.log that looks like "can't create client handle".

We tried to stop nfs and nfsmount and that does not solve anything.

We manage to reproduce the problem, it seems to be a problem that occured when NFS is going through some firewalls. (was said by French support but no solutions).

One solution is to de-activate NFS locking... and my clients are only lock for 15 seconds

Regards

Laurent ROCH