NFS sticks ?

[MoshiachNow]

Hi,

Running AIX 4.2.1 exporting over NFS severall filesystems,occasionally ALL systems mounts to this server gets stuck.

I have checked:
1.Daemons - looks like all are up
2.errpt - nothing
3.stop/startsrc -g tcpip/nfs - no good
4.nfsstat does not show anything special.

ONLY reboot of the server helps !

Any ideas ?

"Long live king Moshiach !"

http://www.7for70.com/

 

[MoshiachNow]

Investigating this problem - I have written some monitoring script.

I'd like this script (extra to ps -ef,etc) to log the nfs status.

Have not found anything better then "nfsstat"

Are there better commands to check for NFS errors/status ?

"Long live king Moshiach !"

http://www.7for70.com/

 

[d3vzero]

Hi Levw,

I would suspect the network.

Try doing the following network checks on both Server and other hosts, prior to the issue and during it, to look for discrepancies.

- ARP Cache.
- Traceroute.
- Ping

Is a firewall involved ?
Are you on Token Ring, or Ethernet ?
Are you using EtherChannel, or just one adapter ?

One other thing to check at the NFS end of things is RPC.

I would have a hard look at portmapper, to make sure it had not been stopped/restarted at some point.

Another command you can use to gather info is rpcinfo. The man page shows some examples of its use.

Hope some of the above helps.

Be interested to hear what the issue was when you nail it ;-)

Best Regards, Peter.

 

[MoshiachNow]

Thanks,

1.No firewall
2.4 Ethernet adapters,static route defined on all interfaces to specific AIX machines
3.How would you use rcpinfo in this specific case ?
(I have a script logging any info there every 60 seconds)

Thanks

"Long live king Moshiach !"

http://www.7for70.com/

 

[MoshiachNow]

Any suggestions on using rpcinfo ?

"Long live king Moshiach !"

http://www.7for70.com/

 

[d3vzero]

Sorry for the delay there ... out sick. :-(

To check the RPC Registered Programs.
rpcinfo -s NFSServerHostName

To Check RPC Statistics. (I would have a close look here).
rpcinfo -m NFSServerHostname

If you are doing it at reqular intervals I would suggest issuing these rpcinfo calls both on the Server, and on a client (but both looking at the server). When running on the Server just skip the rpcinfo Hostname arg.

Are all four adapters used as a transport for NFS from this Server to various clients ? Or just one of them ?

How many clients have you got using the NFS Server ?

Do all Clients lock up at the same time ?

Do you have any other network related services on this NFS Server, accessed by the same Clients, that are impacted once it goes bad on you ? (ie. ping, rsh, telnet, ssh etc ...)

I'm sure you have probably already thought of all these, but just in case you did'nt. ;-)

Check your "no" settings. thewall sb_max send/recv spaces etc.

Make sure you capture snapshots of your ARP Cache. (arp -a)

Also check for MBUF Fails (netstat -m), and also monitor actual packets on the adapters (netstat -I en? 10).

Have a look at netstat -v as well as this will show you total adapter statistics.

Best of Luck, Peter.