From: Udo Grabowski (udo.grabowski@imk.fzk.de)
Date: Wed Jan 15 2003 - 12:17:42 EST
Thanks to Dr.T.Blinn and Jayant Nash for their clear answers.
As suspected this (original question below) cannot be done from
/etc/fstab (only for NFS). So I've written a small script for
/sbin/init.d which is linked in rc3.d and rc0.d (to unmount).
I can send it if someone needs a quick solution.
======== Original question ============
> I have several harddisks in our 8 cluster ES40 with a advfs domain
> Work_<n>#Work (<n> = 1-8) and want to mount them on a
> global directory /Work, which is a cdsl to /cluster/members/{memb}/Work,
> so that every machine has it's own space associated with this directory.
> I can do it by hand 'mount Work_<n>#Work /cluster/members/member<n>/Work',
> but placing 8 such mounts into /etc/fstab does nothing. Mounts with
> local directories seem to be completely ignored by the CFS.
> But this mechanism works when mounting NFS directories this way. And
> how and where are the root<n>_domain disks mounted, which are also not
> mentioned in /etc/fstab ? Where must I place the mounts to get them at
> boot time ?
======== Tom's answer ===============
If I understand your question correctly (and I'm not sure that I do),
you have an 8 member cluster, and you have 8 domains, each of which
has a single fileset. Each domain is intended to be used on just one
member.
I believe you can't do this from /etc/fstab because it's really meant
for things that are highly available. These are not, at least, not
as you describe them.
What I think you want to do is create a script that runs from your
/sbin/init.d, probably from rc2.d, that does the mount on the way up
on each member, using the mount command as you describe. You want
to do it that way so that only that member serves the file system
in the cluster (if in fact it ever gets served).
NFS is just different, as is mounting the cluster root. Forget what
you think you know about how UNIX works, TruCluster isn't UNIX and
it changes all kinds of things.
========== Jayant's answer ==============
I had the same situation, n there are 2 ways to work
around with local disks mounting.
1) If you use the server_only option in the fstab that
would work only for that node, all other entries will
give a small error at boot time, n will continue
booting.
2) This is what i did, i wrote a script which will
mount and unmount the file system stop the volume and
deport the diskgroup.( if under lsm control)and
created links in rc3.d and rc0.d( S99mount_local and
K01unmount_local ).So u will not get a domain panic
message on other nodes when the system goes down.
In the first method u will not be able to see the
mounted dir on other nodes ( as this server_only).
If you want i can give u the script.
About mount rootvol its done using the bcheckrc script
in /sbin, which in turn calls the mountroot script.
-- Dr. Udo Grabowski email: udo.grabowski@imk.fzk.de Institut f. Meteorologie und Klimaforschung II, Forschungszentrum Karslruhe Postfach 3640, D-76021 Karlsruhe, Germany Tel: (+49) 7247 82-6026 http://www.fzk.de/imk/imk2/ame/grabowski/ Fax: " -6141
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:04 EDT