TR Forums

I have a lsi megaraid sas 9271-8i controller with a raid 1 ( two drives) and raid 5 (three drives) and a dedicated hot spare for the raid5. One of the raid 5 drives is kicking up an error in the Pred Fail Count, about 1 more each day. I'd like to replace the drive.

My question is, if I right click the drive I have the option to "replace the drive" or "make drive offline". If I make the drive offline, is the hot spare going to jump in. If that is what happens, do I need to wait for the rebuild to complete before taking the server down to replace the drive? If I choose replace the drive, what will happen?

I'm looking for input on the safest route to take.

If you take the drive offline, the raid 5 will enter a degraded state and start a rebuild of the array using a hot spare. That should be automatic, otherwise your other disk would be labelled a "cold standby", and not a "hot spare". The term hot spare specifically means "available for immediate automatic failover" You should wait for the rebuild and the first integrity scan to finish on the hot spare before doing anything else.

If you use the "replace drive" option, it depends on your controller and what firmware/software you're using but I think with LSI MegaRAID what it does is copies data onto the hot spare first and then when that is complete you should replace the failing disk and it copies the data back from the hot-spare onto the new disk without going through the lengthy rebuild process.

I'd see if there's any more information about the "replace drive" option in the help or user manual of the software you're using. It varies between different versions of MegaRAID so I can't give you a definite answer on that.

You do *not* want to offline the failing drive. That will give you a 2-drive RAID 0 that will then regenerate the parity onto the spare.

If you do a replace, it will utilize all three drives currently in the array to preemptively copy to the spare and then activate it. You can then swap out the failing drive for a new one, and it'll do a quick copyback.