NFS-root mounting problem

[Michael Prinkey]

We ran into a similar problem in building a netboot cluster.  The 
normal startup process is trying to mount the root partition again.
What you really need to do is remount it and make sure it is rw.
You can try that.  I think it might require changing the rc.sysinit 
script.  Or perhaps just take the root listing out of /etc/fstab

FWIW, we found the whole root NFS approach to be more trouble than 
it was worth.  Instead, we built a 4MB ramdisk and used it as /. 
Then we have several ro mounts that go to /bin, /usr, etc. which 
are used for all of the nodes in the cluster and only have rw /tmp 
and /var for each node.

Mike Prinkey
Aeolus Research, Inc.

At Monday, 1 July 2002, Jakob Oestergaard <jakob at unthought.net> wrote:

>Dear 'wolfers,
>
>I'm forwarding this mail to the Beowulf list, it was previously sent to
>linux-kernel - but I guess the most NFS-root experience is on this 
list.
>
>I'm trying to net-boot a system (root NFS) here, but I run into trouble
>mounting NFS filesystems *after* the root filesystem has been mounted
>(over NFS).
>
>Both server and client are 2.4.18 on i686 - everything works great if I
>boot from a disk and mount the home filesystem via. NFS.
>
>Booting the client via. pxelinux, the following happens:
>*) Kernel loads, gets IP, ...
>*) / is mounted (rw) via NFS
>*) Init-scripts start running, brings up eth0, lo, portmapper, syslog,
>NFS locking services, all is ok...
>*) Init-scripts attempt to mount the home filesystem from the server,
>just like it would if booting from the disk. This fails with the
>following errors:
>--------------------------
>Mounting NFS filesystems:  exec: Stale NFS handle
>mount: Stale NFS handle
>exec: Stale NFS handle
>
>touch: creating '/var/lock/subsys/netfs': Stale NFS handle
>
>Mounting other filesystems:  exec: Stale NFS handle
>dup2: Bad file descriptor
>exec: Stale NFS handle
>--------------------------
>
>I tried net-booting to single-user mode, which works perfectly well.
>Then, mounting *any* NFS filesystem will lock the system with stale NFS
>handles again.
>
>The *really* interesting thing is, that after one such failed attempt,
I
>must re-start the NFS *server*, otherwise the client gets a "Root-NFS
>server returned error -13 while mounting /netboot/falcon"...  But I
>suppose this problem will go away with the first one.
>
>Anything I can try ?   Any ideas ?  Need more information ?
>
>-- 
>.................................................................
>:   jakob at unthought.net   : And I see the elder races,         :
>:.........................: putrid forms of man                :
>:   Jakob Østergaard      : See him rise and claim the earth,  :
>:        OZ9ABN           : his downfall is at hand.           :
>:.........................:............{Konkhra}...............:
>_______________________________________________
>Beowulf mailing list, Beowulf at beowulf.org
>To change your subscription (digest mode or unsubscribe) visit http:
//www.
>beowulf.org/mailman/listinfo/beowulf
>