Replacing failedsets HSG80

[Lawrence Bowen]

Hi all 2 questions really...

1) With autospare enabled, am i right in asuming that if i replace the failed disk with a brand new one in the same slot, that i will still have to delete the failed disk

2) Once a failed disk has been replaced, what is the procedure to rebuild the raidset so that the disks are in alligment again from the Command Console point of view?

thanks!!! - lawrence bowen

[Richard Sharwood]

Hi Lawrence,

You are quite right that you will have to delete the failed disk from the failed set.

In order to re-align the disks once the failed disk has been replaced, you need to do the following:

SET (your raidset name) NOPOLICY
SET (your raidset name) REMOVE DISKyyyyy (Where yyyyy is the disk number of the spare disk)
SET (your raidset name) REPLACE DISKxxxxx (Where xxxxx is the disk number of the replaced disk)
SET (your raidset name) POLICY = (your replacement policy)

The spareset disk then moves into the failedset and needs to be moved back.

DELETE FAILEDSET DISKyyyyy (Where yyyyy is the disk number of the spare disk)
ADD SPARESET DISKyyyyy (Where yyyyy is the disk number of the spare disk)


Hope this helps


Richard

[Lawrence Bowen]

Thanks that is great, i tried it and it worked spot on!

-lawrence

[Spike Burkhardt]

Lawrence & Richard,

This seems to complicated a procedure. Maybe it needs to be because of the command console software. We don't use it where I work. In our environment, we have two disks in our spareset. We have autospare enabled on failedset. When a disk goes south, the failed disk gets put in the failedset storageset. One of the disks in the spareset then gets swapped into the raidset that had the failed disk if it meets the replacement criterion(size of disk). All this happens automatically.

If you don't have a disk in the spareset then I guess you might need to do Richard's procedure, I'm not sure. You will need to do some command line manipulation if you put the new disk into the slot where the failed disk was.

Spike

[Lawrence Bowen]

Hi Spike, i hear you, but this will eventually result in all the members of the set being out of Target aligment. The complicated steps above are about how to re-align them.

-lawrence

[Mike Naime]

Spike:

I understand exactly what Richard and Lawrence are saying.


After the Autosparing that you are talking about has occured, you may want to reset your raidset so that it has the original members in it. Of so, then you need to do something similar to what Richard outlined.

Example: If my original raidset was composed of DISK10500 DISK20500 DISK30500 DISK40500 DISK50500 DISK60500, and DISK30500 is failed with DISK31500. I may want to get the raidset back to its original configuration of using row 05 across all 6 channels. It's not necessary, but it makes troubleshooting at a later date easier.

[Christian Schwartz-Søre]

What exactly are you refering to when talking about the members of a raidset being out of target alignment, and they need to be realigned

regards
Christian

[Richard Sharwood]

Christian.

If you have a raisdet that is split across the 6 disk shelves and one of the members fail and the sparedisk is used, it is possible that this spare disk was on the same shelf as other raidset members therfore having 2 disk from the same raiset on the same shelf.

This would then leave you in a vunerable state, if the disk shelf failed you would then lose 2 disks from the raidset which cannot be recovered from.


Many Thanks

Richard

[Mike Naime]

Spike:

Another point -- I have 6 disks in the spareset of all of my HSG's. One from each channel (Shelf) of the storage array. This way, if I have a drive in channel 2 fail out, I can replace it with a drive from channel 2.

In your situation. If you had a raidset that was composed of members from channels 1, 2, and 3. and your spareset drives where in channels 1 and 2. If you lost the drive in channel 3, you are putting yourself at risk if it is replaced by a spare from either channels 1 or 2.

I have had a channel/bus go bad 2 times in the past 3 years.