Subject: Re: Problems with TCP-based NFS mounts hanging
To: Frank van der Linden <firstname.lastname@example.org>
From: Jan-Hinrich Fessel <email@example.com>
Date: 06/26/1997 23:49:07
Content-Type: text/plain; charset=us-ascii
In message <199706262052.WAA11358@mail.wins.uva.nl>you write:
> Quoting Greg Wohletz,
> > Yep, I've seen this. Also I have found one major disadvantage to tcp
> > mounts is that they don't deal with nfs server reboots, if the nfs
> > server goes down and comes back up a new tcp stream is not established and
> > the client basically has to be rebooted in order to continue using the
> > nfs filesystem.
> > I've also seen random hangs over tcp mounts even when the fileserver didn't
> > reboot.
> When I originally tested the NFSv3 code against Solaris servers, I saw
> this once, but could not reproduce it, and neither could others. But
> apparently there still is a problem. If you could me provide with as much
> information as possible about your environment (client, NetBSD version,
> server, OS version, mount arguments, possible kernel messages, when
> does it happen, has the client been idle for a while, etc), I will try my
> best to fix it.
I for my part can reproduce it when doing a make build on a nfs-mounted
/usr/src partition. But, sorry I can't really do that because my client part
of the testbed is now in production. However, both client and server are
quite current i386's, and, if that will help, nfs over tcp continues to work
after killing all nfsd's on the server and starting up new ones. So, it might
not at all be client-related...
Tragbar ist, was nicht herunterfaellt.
-----BEGIN PGP MESSAGE-----
-----END PGP MESSAGE-----