Apache OpenOffice (AOO) Bugzilla – Issue 25824
OOo user setup fails on shared NFS root
Last modified: 2004-02-24 14:55:52 UTC
Hi :) I'm using OOo from Debian sid, and seem to have the same issue as this guy http://lists.debian.org/debian-openoffice/2003/debian-openoffice-200309/msg00075.html I have followed bugreports through http://www.openoffice.org/issues/show_bug.cgi?id=17517, but I already use nfs-kernel, and the STAR_PROFILE_LOCKING_DISABLED=1 export STAR_PROFILE_LOCKING_DISABLED suggestion made no difference. In the end I mounted the rootfs with the 'nolock' option and this has cleared the problem, but I am glad that the rootfs is mounted readonly because I definately wouldn't have followed this path were it rw! The Setup_err.txt from the client said "can't open registry file <fullpath>/types.rdb" even though that path and file exists and is world readable :/ Cheers, Gavin.
of @ mhu: Any suggestions?
Hi Olaf, Gavin, Usually (see e.g. issue 19339) this is a matter of system 'misconfiguration', i.e. the NFS client (and possible the server as well) does not run a NFS 'lockd' process. Depending on kernel and system configuration, this can either be builtin '[lockd]' (as shown by 'ps') (kernel configured with CONFIG_LOCKD*=y) or an external '/sbin/rpc.lockd' (a 'friend' of '/sbin/rpc.statd'). On my Debian Woody, all this is taken care of in '/etc/init.d/nfs-common'. Hope that helps, Matthias
Resolving as 'works for me'...
Ah cool. That uncovered quite a subtle error... As I PXE-boot, I use an initrd to set the system up for the NFS-root.. and as part of that, it runs a statically-linked 'mount' before the real 'init' gets going. So, when the real 'init' takes over, the NFS-root *is* mounted fine, but isn't listed in the output of a 'mount' command, and hence the kernel [lockd] does not run. I added the nfsroot into the /etc/fstab of the initrd image (which gets used in favour of the NFSroot one...) and the exports get re-mounted later in the boot sequence. All is well :) Cheers, Gavin.