Sun Cluster 3.0 - Rearranging disks

Sun Cluster 3.0 - Rearranging disks

Post by Nick Havar » Thu, 25 Dec 2003 00:55:21


Due to a number of problems with my SVM disk groups, I am thinking about
rebuilding them. Could you please let me know if I can do the following.
Assume the data will be backed up, so it doesn't matter if I lose the
data/partitions etc.

The system is a small two node Sun Cluster 3.0 system, where only the shared
disks are under the control of SVM (i.e. /, /opt, /tmp, /var etc are all
/dev/dsk devices, an not /dev/md devices).

To wipe out all disk groups etc, is the easiest way is to delete the
metadatabases on each node?
metadb -f -d <metadb device>

Will this remove all my configured disk groups etc. I have one that I am not
able to take control of and subsequently can't delete. Presumably this will
get rid of it once and for all.

My cluster has two Sun 711 Multipacks for shared storage. If I moved the
disks around in each pack, and between packs will Cluster 3.0 be able to
remap all the disks etc, back to the same did device number, if not what do
I need to do t get SC3 to rebuild the did list. I wish to do this becuase
the disks are all over the place in the disk arrays as I've bought the and
expanded the capacity over several months. Essentially I want to put things
in to a reasonable order etc.

Then I can rebuild the necessary meta databases using metadb on each node,
and the groups as appropriate etc, and put back the data to each device?

Are there any big risks to my cluster in doing the above, that I am
Thanks for your help.


P.S. Remove xyz to reply

Sun Cluster 3.0 - Rearranging disks

Post by dejanew » Fri, 16 Jan 2004 05:29:13

Backups are your friend, and it's good you have them. The command you
gave will clear the meta database, but there are a number of config
files in /etc/lvm. You'll also notice some config in /etc/system
depending on how far you want to back Disksuite out. To rebuild the
DID, man scdidadm. Chances are you're going to lose your data in this
process unless you're careful. From the amount of work it sounds like
you want to do though, you may be better off just doing a clean
install. Reloading the OS and reinstalling Cluster would only add a
day or so at most to your list of tasks, and then you'd have a known
baseline. Also, you should probably think about putting your root
volumes under Disksuite control if they're not hardware mirrored. The
entire purpose of a cluster is HA, and I'd hate for you to have
availibility issues because of a failed system disk.