Re: NFS 4 error

[Malahal Naineni <malahal@us.ibm.com>]

J. Bruce Fields [bfields@fieldses.org] wrote:
> On Wed, Sep 04, 2013 at 11:20:45AM -0500, Malahal Naineni wrote:
> > jupiter [jupiter.hce@gmail.com] wrote:
> > > Hi,
> > > 
> > > I am running NFS 4 server on CentOS 6.4 for many months, it works
> > > well. Recently I found that the dmesg display many following errors
> > > but it still works. I searched google but could not find result of how
> > > to fix it. What are those error message means? Are the errors serious
> > > or not? How to fix it?
> > > 
> > > --------
> > > rpc-srv/tcp: nfsd: got error -32 when sending 24 bytes - shutting down socket
> > > rpc-srv/tcp: nfsd: got error -32 when sending 24 bytes - shutting down socket
> > 
> > When the nfsd (NFS server thread) was trying to send something, it got
> > -32 (EPIPE) error.  I have seen this kind of errors when the socket is
> > closed by clients.  Clients do close the socket (and recreate one) when
> > they need to retry due to network failures.
> > 
> > They are NOT serious errors, and if they are still happening, then look
> > at your network errors/retries etc.
> 
> We should really just demote this to a dprintk, in that case.
> 
> The network failures are potentially a serious problem if they mean the
> client's forced to retry a nonidempotent operation.  But we probably
> shouldn't be spamming the logs by default?

I haven't seen many of these messages. That is the only message logged when
the socket gets closed. It is a KERN_NOTICE level message. We can rate
limit it, but in my experience, I haven't seen a need for rate limiting
it.

Regards, Malahal.