On 14/05/19 18:41, Eric Valette wrote: > On 14/05/2019 19:25, Julien ROBIN wrote: > > Thanks a lot. I think I'm ok (rather will be in 5 hours). > >> mdadm --assemble /dev/md0 /dev/loop0 /dev/loop1 /dev/loop3 /dev/loop4 >> mdadm: /dev/md0 assembled from 3 drives and 1 spare - need all 4 to >> start it (use --run to insist). > > I only did mdadm --assemble --run /dev/md0 > > and it started rebuilding automatically using the spare. Dunno why it > did not work automatically (did not put --run the first time I tried). If an array is degraded for some reason, it will often refuse to start running. It actually makes a lot of sense that the array should refuse to come back without some sort of operator intervention. (That refusal is only for the first time - if the operator starts a faulty array, it will then happily re-start.) This is (sort of) documented on the raid wiki. I'd love to add more detail but my raid-fu isn't good enough to know what actually happens, so I can't document it ... :-) Cheers, Wol
