Here's the scenario:
We currently have a lun that is about 200G raw (150G usable) that is filling up fast and we need to expand -
The options I can think of are:
Create another large lun (I'm shooting for 400G) and migrate
Create a new meta-lun with 2 200GB luns on 2 different raid-5 groups and migrate
or Create a new 200GB Lun on a raid-5 group not currently part of the existing lun and concatenate into a meta-LUN with the existing LUN.
Then after expanding, go into disk mgr on the server and expand the partition to see the space.
Are there any clear winners out of these options? The LUN isn't hit really hard as far as high IO rates, it's just big. Does any of the 3 involve down time for the LUN? Are there any Gotchas or Look-Out-Fors?
also, assuming option 3, is it best to make the new LUN the same size as the old LUN
and finally, I seem to vaguely remember that it's best to make any lun an even division of some parameter to maximize performance. Does anyone remember how to calculate this?
I would go with option 1; create another large lun (e.g. 400GB) and migrate to it.
This is the simplest and easiest to manage and keep track off.
If opting for using a MetaLUN construction, when working in one raid group I'd recommend concatenated metalun. When going across identical raid groups (same number of drives, same drive spindle speeds, same raid type), a striped metalun is probably a better choice for performance.
If you don't need the performance, and aren't limited by the available space (to where you cannot find a raidgroup with enough space to fully migrate to), I'd go for the simple end solution.
If you're running 2003, after you finished the LUN resize (whichever option you end up chosing), verify in the disk manager that the new size is seen, but to resize the partition, use diskpart following the steps in this MS KB article.
Member since 2003
Rats - I realized as i was reading your post that we don't have a large enough block to do option 1 - so it looks like its striped metalun... how disruptive is striping process to the user - should I hold off until the weekend?
Striping won't take any downtime, but there will be a performance hit, so doing it in the weekend it probably a good idea (especially as there's no cancelling/stopping once you start it).
Member since 2003