https://bugzilla.novell.com/show_bug.cgi?id=415607
User nfbrown@novell.com added comment
https://bugzilla.novell.com/show_bug.cgi?id=415607#c19
Neil Brown changed:
What |Removed |Added
----------------------------------------------------------------------------
Info Provider|cbenson@fhcrc.org |R.Vickers@cs.rhul.ac.uk
--- Comment #19 from Neil Brown 2008-09-24 23:50:38 MDT ---
Thanks for the traces. I can now see some of what is happening.
The tcpdump trace shows normal NFS startup handshake where the client
first checks with portmap to find mountd then talks to mountd to get
the root filehandle. Then it checks with portmap again to get the
port for NFS (2049) and then connects to the NFS server.
You would expect to then see an FSINFO call, an FSINFO reply, then a
PATHCONF call and a PATHCONF reply.
Instead we see various other calls such as GETATTR, FSSTAT, PATHCONF, all
for *other* filehandles. There calls are made consecutively without waiting
for a reply. As soon as the server sees the PATHCONF request (with a
different filehandle to what the mount returned), it closes the TCP connection.
The client then opens the connection again submits several requests one of
which fails.
What must be happening here is that several mounts are all being attempted
against the same server at the same time. When this happens the client will
use the one TCP connection for all of the mount points. The PATHINFO
for one of these is causes the server to (incorrectly I think) close the
connection. This just causes it to be retried again and again making
no progress and blocking any other mounts from the same fileserver.
So it seems like the NetApp filer is doing the wrong thing, closing the
connection.
However Bob's situation doesn't involve a NetApp filer.
So maybe Bob's problem is completely different and just looks similar,
or maybe I'm missing something.
Bob: are you able to collect a tcpdump trace of traffic when the problem
occurs?? Thank might be helpful.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.