booting live system from nfs-user-server fails at nfsmount

Bug #189593 reported by chrysn on 2008-02-06
14
This bug affects 1 person
Affects Status Importance Assigned to Milestone
nfs-user-server (Ubuntu)
Undecided
Unassigned

Bug Description

when booting the live cd via pxelinux (tftp) and nfs (using nfs-user-server), mounting the nfs share fails with

+ nfsmount -o nolock -o ro 10.0.0.227:/var/lib/nfsboot/ubuntu /cdrom
rpc failed: 2

this can be fixed by replacing nfsmount with normal mount in initrd.gz's scripts/casper at line 215 (gutsy and hardy). the problem exists at least since feisty and is still present in hardy alpha 4.

i've attached the casper.log file generated.

in case it is important, you can look up details of the process at [1], writing which i discovered the bug.

updated 2008-02-29: it turned out this only happens with nfs-user-server, so if it's a server side issue, someone who can judge nfs problems please close this bug or move it to nfs-user-server.

[1] http://christian.amsuess.com/tutorials/lanbootserver/#previous-problems-with-nfs

chrysn (chrysn) wrote :
chrysn (chrysn) on 2008-02-29
description: updated
Mantas Kriaučiūnas (mantas) wrote :

chrysn wrote at 2008-02-29:
> it turned out this only happens with nfs-user-server, so if it's a server side issue, someone who can judge nfs problems please close this bug or move it to nfs-user-server.

I'm not 100% sure, that this is only nfs-user-server issue, maybe this problem is in both packages - nfsmount and nfs-user-server.
I'm assigning this bug to nfs-user-server, let's wait for answer from nfs-user-server maintainers or developers :)

Nico (pub-1g6) wrote :

I found that replacing nfsmount by mount in the initrd got it to work,
(file scripts/casper line 209), got nfs-user-server working with live booting on nfs, so this a bug in nfsmount.

Peter Cordes (peter-cordes) wrote :

I tried mount from an initramfs, while trying to boot the livecd with nfsroot. Regular mount works on Hardy, but the kernel prints warnings that it's providing a workaround for not having a lockd running. This support is no longer present in Intrepid's 2.6.27.

In Hardy, see linux-2.6.24/fs/nfs/client.c:nfs_start_lockd(): There's a workaround for failing to start lockd, and a big comment about it.
...
printk(KERN_ERR "nfs: Starting lockd failed (do you have nfs-common installed?).
...

 Maybe this is ok if you use the right NFS mount options, like no locking, or something. I was just trying to run mount manually with no mount options.

chrysn (chrysn) wrote :

that warning appears only when the "-o nolock" option is omitted. using nolock will make the warning go away and probably also make it work for intrepid.
that option is consistently being used on the initrd scripts. (although at the moment i could only verify that for debian; makes no great difference i guess)

Confirming this bug because it happens to several users.

Changed in nfs-user-server (Ubuntu):
status: New → Confirmed
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers

Bug attachments