[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#632074: linux-2.6: NFS4 client sends NULL calls in the TCP session



On 24/11/2011 06:31, Jonathan Nieder wrote:
Hi Franck,

Franck Eyraud wrote:
On 29/06/2011 15:38, Bastian Blank wrote:
linux-nfs@vger.kernel.org.
I will ask them also.
Did you get in touch with linux-nfs@?  If so, do you have the
date and subject or message-id of a message so we can track
the discussion and conclusion?

Thanks,
Jonathan

Hi Jonathan,

Thank you for your message.
I'm sorry I didn't give further more notice. The answer from linux-nfs list was more or less that the problem was on NetApp OnTap side, maybe logging too many messages.

Here is the thread where I got in touch with Thomas Haynes and Trond Myklebust from NetApp.
http://article.gmane.org/gmane.linux.nfs/41786

They kindly offered me to analyze the traces I already sent to the NetApp Technical Support.

Here is what Thomas Haynes wrote to me (off-list):
-------------------------
Recapping what I see in the case notes:

From the case notes, it appears the filer is objecting to the v4 NULL probes
because the GSS context is no longer valid.

My guess is that we don't even know it is a NULL probe at this point and
are kicked out at a higher level. Even if I am wrong, we probably need to
process the context in order to construct a reply.

The client changed behaviour from Debian 4 to both Debian 5 and Debian 6.
I have no clue what changed in there.

It appears that the customer doesn't mind this occurring and would be happy
if we could dial down the number of messages logged? I.e., it appears too
chatty?

I don't see in the notes that a request was made to reduce the log messages.
I.e., customer support probably focused on v4 NULL probes.

If reducing the cadence of the messages would help, I can file a bug
on this. Note, we do occasionally want to see these as there may
be other times when this legitimately occurs.
--------------------------

We recently tested with a linux (ubuntu) with the linux kernel 3.0 and the problem still appears. In the mean time, the OnTap server software has been upgraded to version 7.3.6P1.

So we basically still not really sure where this comes from (client or server ?), and we are still experiencing this problem (it seems the bug they opened at netapp side isn't solved yet). I don't have the competences to analyze further the problem, but I can provide traces if someone wants to look into it. It could be also a kerberos 5 issue...

Hope that helps,

Franck Eyraud




Reply to: