Is scrub necessary after sequential resilver if I start rebalancing immediatly for draid?

[littlewu2508]

The usual reconstruction procedures for draid documented in https://openzfs.github.io/openzfs-docs/Basic%20Concepts/dRAID%20Howto.html is

1. Sequential resilver to distributed hot spare (triggered by ZED once disk failure detected)
2. Scrub (automatically launched after sequential resilver)
3. Rebalance (healing resilver) to replace hot spare with new disk

Since rebalance also performs checksum verification of reconstructed data stored in distributed hot spare, is it OK to skip scrub and directly jump to healing resilver to new disk after sequential resilver to hot spare?

[amotin]

I'd say it should be OK to skip the scrub to start rebalance if the replacement disk is available before the scrub completes, since rebalance should do the same checks. But either one or another should better be done ASAP, since otherwise you don't know whether the original sequential resilver did the work right, restoring actual data redundancy, or only made it look so. Especially if the vdev has more redundancy (draid2/3 or multiple copies of metadata), where ZFS might redo things better if the error is found.