*** Copy of a post over at Petri forums ***
Hi,
I am part way through (well most if you consider the amount of data I’ve copied) a data migration from three drives to 3 three others. The drives are SAN attached storage to a couple of blades that form one of the company’s SQL servers.
The current situation is as follows:
Server #1:
New drives available (added as additional drives, formatted & data copied) i.e. it currently has BOTH old and new drives attached. SQL is disabled but clustering is enabled.
Server #2:
Powered down, clustering service disabled.
Now what I have to do is to switch the old drives with the new drives in terms of drive letters and disk signatures (I have the information needed for the signatures) and then get server #2 to recognise the new drives and fail over if required.
My plan (and I’m not sure this is a good one) is as follows:
-
[*]SERVER #1: disable clustering, switch drive letters (so that the new drives become the same letter/order as the old drives only bigger) & write disk signatures to them. Power down.
[*]SERVER #2: Power up. Enable clustering and restart. Ensure cluster forms OK and check drive letters, sizes, order (may require some editing).
[*]SERVER #1: Power up and ensure cluster forms OK (check drive letters, sizes, order etc.).
[*]SERVER #1/2: Check that groups can fail over between cluster members.
[*]Re-enable monitoring and check all systems appear to be functioning AOK.
My first concern is that SERVER #2 might not see the new drives as anything other than RAW drives (they are attached) or may even wipe the data I took so long to copy over (this happened earlier but both servers were up and I disabled the clustering on #1 so the fail over happened live and trashed what I’d done … fortunately not much had happened at that point).
I’ve been told that clustering requires the disk signatures to be precise (that that’s how it recognises the targets) s o my second concern is disk signatures … can I give 2 drives on the same system the same disk signature i.e. are they supposed to be unique?
My biggest concern, of course, is that I’m no clustering expert (our resident is on paternity leave) and that I’ll destroy the company’s data: sure regular backups are done (daily, I started this Saturday so the Friday backup would run and checked that the backup ID wasn’t logged on before I started) but I wasn’t able to check remotely.
Anyway, any advice sorely appreciated.
Kyu
