illumos gate

Hi all

I have a server with four RAIDz2 VDEVs, and the other day, a drive, c14t19d0, died. The spare, c9t7d0, stepped in and resilver started. I did my zpool offline/cfgadm -c unconfigure/zpool replace/zpool detach <spare> and resilver now runs on the new c14t19d0. This was all fine, until a second drive, c14t12d0, in that VDEV died. This made the spare, c9t7d0, take over again and force a /restart/ of the whole resilver process without regard to the current resilver, which was almost finished. zpool status now shows the resilver process as just started, and ETA some time in the future.

Now, this rises a few issues. If the resilver process had continued until the initial disk was finished, I'd have better redundancy while waiting for the second disk to resilver, and could potentially lose a third disk during that time.

For the record, this pool has been scrubbed periodically, so it's just another of those things that happens when you don't have time for it.

In my opinion, this is a bug or a design failure, and it would be good to get it fixed.

Best regards

roy