AutoRAID 12H - to rebuild or not?

[Denver Osborn]

Hey All,

We had a disk on one of our 12H's fail today and auto rebuild is enabled and we had hot spare active. The rebuild had a high priority by default... the rebuild started a just before 8am, so the box slowed down a lot when users started getting on. Our apps performance began having problems, and I assume due to the high i/o from the rebuild. Lowering the priority of the rebuild helped us out.

What's everyones opinion on auto rebuild? Do you have it enabled or not? I've already mentioned to the higher ups that it all depends on how important the data is.

Another question is, let's say we disable auto rebuild. After we replace a failed disk and it's included, we'd then initiate an arrayrbld. What's different between the manual and auto rebuild? Will the later not rebalance, just populate the replaced disk with the data.

Thanks for your input!
-denver

[A. Clay Stephenson]

My preference is Active Hot Spare enabled and Auto Rebuild ON BUT with Rebuild Priority set to LO so that server I/O takes precedence over array rebuilds. If you don't have Auto Rebuild enabled AND a second drive fails then you are toast. I can imagine a second drive failing over a long holiday weekend. I only want to imagine that not live through it! Auto Rebuild enabled starts the rebuild immediately upon disk failure so that your window of vulnerability is lessened. By setting your Rebuild Priority to LO, you have dramatically decreased the performance hit with only very moderate increase in risk to the data.

[Denver Osborn]

I think I'll recommend we keep it on but with a low rebuild priority. Hopefully it won't kill the app if/when another disk fails.

In reality we could disable auto rebuild because we mirror the data w/ lvm to another 12h (this is done because of a prior 'close call' w/ a hw fault on the 12h)

Would you happen to know how the rebuild works if I replace the failed disk, then manually issue an arrayrbld command? I'm thinking it should only write data to the replacement disk and not the entire array... oh-well, time to pull out a manual when I get to work in the morning. :)

Thanks for the feedback Clay.
-denver

[Adey]

The Autoraid 12H doesn't work on a disk basis.

To have active hot spare you need to have space on the whole of the array equal or greater than that of the largest disk you have in the array.

Ie. If you have all 18GB disks then you need to reserve at least 18GB of disk space in the array to allow it to utilise hot spare capability.

If you notice when a disk fails and it is replaced it performs a balance rather than a re-build.