On Wed, Jun 26, 2013 at 5:36 PM, Tom H <[log in to unmask]> wrote:
> On Wed, Jun 26, 2013 at 2:24 PM, Eve V. E. Kovacs <[log in to unmask]> wrote:
>>
>> We have an SL5.5 nfs server that has developed an odd problem.
>> We have configured /etc/sysconfig/nfs to assign port numbers for the various
>> services, 662 for statd and 892 for mountd, in particular.
>> For reasons unknown, rpc.statd, in addition to running on port 662 as
>> directed, has grabbed port 892 for running udp.
>>
>> We see, on the server:
>> rpc.statd 2412 rpcuser    3u  IPv4   7443       UDP *:662
>> rpc.statd 2412 rpcuser    6u  IPv4   7434       UDP *:892
>> rpc.statd 2412 rpcuser    7u  IPv4   7446       TCP *:662 (LISTEN)
>>
>> Since 892 was the port that was assigned to mountd, it caused mountd to
>> fail, and hence nfs mounts from the clients to fail.
>>
>> We have switched mountd to run on port 895 for the time being, so we are
>> functional, but we would like to understand what happened.
>> We are running an identical backup server, and interestingly, there, the
>> extra port being grabbed by statd is 982 not 892!
>>
>> Has anyone else seen this behavior, or have any idea why the two servers are
>> behaving differently?
>> Presumably this extra port for statd is being assigned by portmap.
>> Is there any way to fix this port assigment so that we don't get a collision
>> like this in the future?
>
> Please post the NFS server's "/etc/sysconfig/nfs".

And strongly consider updating to 5.9. There are ongoing enhancements
of both security and critical services under load in the kernel,  in
the automount utilities, and potentially in the NFS tools.