On Sun, Nov 19, 2017 at 4:00 PM, Larkin Lowrey [off-list ref] wrote:

I did a 'pull-the-drive' test on a raid6 array with a write-journal and
found that mdadm (v4.0-218-gfe05dc4 - 2017-11-09) would not let me --re-add
the drive so I add to --add it.

mdadm: --re-add for /dev/sdq1 to /dev/md124 is not possible

My expectation was that the drive would be recognized as being the missing
member and would be re-added and only out-of-sync regions would be resync'd.

When I --add'ed the drive to the array the resync process started for the
entire array, not just for the out-of-sync regions. Is this correct behavior
for arrays with write journals? Are there plans to perform a selective
re-sync rather than a full re-sync?

Needs to be in the mdadm.conf for this array, see 'man mdadm.conf'

-- 
Chris Murphy