Hmm, maybe some race condition that sometimes breaks --re-add immediately after --remove? Might be worth trying separating it into two commands, i.e.: mdadm

Hmm, maybe some race condition that sometimes breaks --re-add immediately after --remove? Might be worth trying separating it into two commands, i.e.: mdadm /dev/md0 --fail /dev/sdb1 --remove /dev/sdb1 mdadm /dev/md0 --re-add /dev/sdb1 --update=no-bbl I have a simple RAID-1 with metadata version 1.2, too. My testing was done on Debian unstable (Linux v5.8, mdadm v4.1).