NFS client issue

[Jason Joines]

Net Llama! wrote:
> On Fri, 26 Jan 2007, Jason Joines wrote:
>> Net Llama! wrote:
>>> On Fri, 26 Jan 2007, Jason Joines wrote:
>>>> Net Llama! wrote:
>>>>> On Fri, 26 Jan 2007, Jason Joines wrote:
>>>>>
>>>>>>    I have a bunch of boxes backing themselves up to an NFS server.  One
>>>>>> of them suddenly stopped working, it can no longer mount the NFS
>>>>>> export.  Doesn't seem to be an issue with the server as all the other
>>>>>> clients are still working just fine.
>>>>>>
>>>>>>    The failing client is an old SuSE 8.1 box with a 2.4.19 kernel and
>>>>>> the server is OpenSuSE 10.0 with a 2.6.13 kernel.  The output of the
>>>>>> failed backup script reports "mount: RPC: Port mapper failure - RPC:
>>>>>> Timed out".  portmap is running on the client and I've restarted it but
>>>>>> that didn't help.
>>>>>>    The other clients are all SuSE with versions ranging from 8.0 to 10.0.
>>>>>>
>>>>>>    I tried to do the mount manually a few times ( same server but
>>>>>> different exports and local mount points ) and each resulted in a hung
>>>>>> mount process.  The int option was used each time but none of the mount
>>>>>> processes can be killed even those with the soft option.  The mount
>>>>>> called from the script does not result in a hang, it just fails.  Only
>>>>>> the manual attempts hang and have now been hung for over 18 hours.
>>>>>>
>>>>>>    Using rpcinfo, the client can list the services registered with RPC
>>>>>> on itself and on the server including portmapper, nfs, and mountd.  I'm
>>>>>> using TCP.
>>>>>>    During one of the mount attempts I logged all the network traffic
>>>>>> between the client and the server.  The client contacted portmap on the
>>>>>> server and got replies, it then contacted mountd on the server and got
>>>>>> replies.  However, it never sent anything to nfs on the server.
>>>>> What does this return from the broken system?
>>>>> showmount -e $servername
>>>>>
>>>>
>>>>     Thanks, I hadn't run across showmount before.  In this case it
>>>> showed all the exports on the server along with a list of all the hosts
>>>> allowed to mount it including the host name of the broken client.  The
>>>> output was identical on a working client.
>>>>
>>>>     Also, the server does log the mount attempt from the broken client:
>>>>
>>>> Jan 26 08:19:54 siva rpc.mountd: authenticated mount request from
>>>> broken-client:885 for /local/backup/weekly (/local/backup/weekly).
>>>>
>>>>     It's just the client then never tries to connect to nfs on the server.
>>> Oh the irony.  I ran into this mess just yesterday, and it turned out to
>>> be some kind of NFS server side horkage that went away only after
>>> rebooting.  I'm not sure that helps with your problem though.  sorry.
>>>
>>
>>
>>     Did rebooting the server fix your problem or did you have to reboot
>> the client?  In my case rebooting the server is no problem as it's only
>> purpose is for backups, none are scheduled until later tonight and I can
>> reboot it remotely.  However, the client is being used by lots of folks
>> and will be more problematic to have to reboot.
> 
> Rebooting the server fixed it.
> 



     Rebooting the server didn't fix it for me.  I eventually had to 
reboot the client and that did the trick.  Still have no idea what 
happened though.

     I sure to hate resorting to reboots to try and resolve an issue. 
Makes me feel like a winblows user.


Jason
===========