topic Re: NFS write error in Operating System - HP-UX

NFS write error

T.S.SHIVAKUMAR — Thu, 31 May 2001 07:41:39 GMT

Hi All,
I have a Sun E450 server where users with home directories on an HPB1000 m/c log in. The setup has NIS, DNS and NFS.
I notice that the "/var/adm/messages" file is getting piled up with the foll error:
May 31 13:07:03 condor unix: NFS write error on host berkeley: error 145.
May 31 13:07:03 condor unix: (file handle: 40000004 ffffffff a0000 b60b4 5 a0000 2 0)
May 31 13:07:03 condor unix: NFS write failed for server berkeley: error 5 (RPC: Timed out)
May 31 13:07:03 condor unix: NFS write error on host berkeley: error 145.
The user home directories are mounted via NFS automounts; I do not face any problem doing NFS writes though the "messages" file keeps piling up. Sometimes the errors may be displayed on the user terminal.
Could you suggest a solution?
Thanks and regards,
Shivakumar

Re: NFS write error

Printaporn_1 — Thu, 31 May 2001 08:27:51 GMT

error 5 is I/O error , may be at that time the NFS server is not available or got network problem.

Re: NFS write error

T.S.SHIVAKUMAR — Thu, 31 May 2001 10:32:40 GMT

Hi !
Thanks for the response. But the error messages are continous causing the root file system to get full if I do not keep tab. The root just has about 90 MB free and the error messages are continous. The error no. is 145 and not 5. Should that give a clue? The network is perfect.
Maybe there is some patch??
Regards,
Shivakumar

Re: NFS write error

Jim Hendrick — Thu, 31 May 2001 15:48:14 GMT

There are a few areas to look:

is it always the same filehandle? It is possible that there was an interrupted system call (due to a server crash/reboot) and the client still has the old file handle.

Are you *sure* the network is not an issue?
(try running nfsstat -c on the client and look at the number of retransmissions. If it is over 5% of total calls, this indicates a problem with the transmission. Next ,zero out the counters with nfsstat -z then let it run a while, then re-run nfsstat -c.)

Look at system load on the server (and client). The server may not run enough nfsd processes if it is a very busy NFS box. The client might need more biod daemons if it is unable to keep up with the I/O on ot's end.

Just some places to look...

Jim