UNSOLVED
Customer wants to 'simulate' failed disk scenario on MD3420 SAN by pulling out a disk
Dear,
We are in the process of delivering a brandnew Server system running on a MD3420.
Everything is running and configured, tested, and we have backups (not in production yet).
The SAN is configured with 2 disk groups, each group is 8 disks in RAID10.
We have 2 'hot spare' disks, these are 'hot spare' for both disk groups.
The customer now wants us to demonstrate the 'RAID10' capabilities. He wants to pull out one of the disks and see that the 'hot spare' will take over.
After that, the customer want to plug the extracted disk back in, and see how the SAN respons to that (will that become the 'running' disk again, or will it become the new 'hot spare'.
Is simple pulling out a disk from a running system a good simulation for a failed disk? Is all of this a good idea? What to expect?
Thank you for your answers
I'll defer to the experts here but I believe that most things on the powervault MD line will behave like this: Removing a disk will flag it as "removed", not failed, but it should still recognize that the RAID is in a non-optimal state at this point, and call in a hot spare. Once the hot spare has finished taking over, and the "removed" drive is replaced, I think that the new drive in the original slot will once again become part of the array, and the hot spare that had taken over will go back to being a hot spare.
Another thing to note, simply pulling and putting a drive back in will often flag it as having a foreign config. However, I think MD SANs are smart enough to circumvent this and just use the new disk, regardless of where it's been before.
Have fun!
Joris85
17 Posts
362
0
Posted September 24th, 2015 13:00
I hope the systems are 'smart enough'... if they are, the only major concern I still have for tomorrow is:
They want to do this test 2 hours before the servers are shipped. If this pulling a disk out will start 'rebuilding' the mirror on a hotspare disk, and it takes longer then 2 hours (600GB 10K disks, RAID10), then we are 'screwed' and I will need to unpower the SAN while it is rebuilding the array.
I believe because of that reason it is best we try to postphone the test until after shipping?