In general, this recommended approach with --raid-devices=3 will work, but will fail during boot too. When 1 from 3 is lost the md array will fall to the degraded mode.
чт, 21 мая 2026 г. в 14:19, Michael Tokarev <[email protected]>: > 21.05.2026 12:15, Dmitry Smirnov wrote: > > > Consider configuration where `/home` partition is on 3 HDDs, arranged > > into mdRAID-1, with a hot spare. Once HDD dies/disappears, a (hot) > > spare is immediately used to replace bad/missing disk and restore > > 200% data redundancy. That is what's expected from such array and > > it is how things used to work on Debian few releases ago. > > I wont comment on the bug report as a whole, but want to comment about > this particular scenario. > > If you have 3 HDDs for RAD1, it's much better and reliable to use > 3-way raid1 from the beginning, instead of 2-way raid1 + hot spare. > This way you eliminate the thin ice in here: if one drive dies and > you'll have just one copy left, the probability to encounter second > failure (which is now fatal!) increases significantly, since during > recovery, *whole* remaining drive has to be read, including areas > which hasn't been touched for long, and where you might face some > bad sectors. And recovering from *that* situation is significantly > more difficult. > > When you've 3-way raid1, symmetrical, instead, everything works as > it should be. And as a bonus, you have better read performance (but > at a price of very slightly worse write performance). > > Also, none of the disasterous scenarious you outlined, wont occur. > > FWIW, > > /mjt > -- Yours truly, Dmitry Menshikov http://dmenshikov.com
