Dnia 2010-01-02, sob o godzinie 21:36 +0300, Majed B. pisze: > From the error mdadm reported, the array's been converted to a 5-disk > one and it's trying to assemble it with 2 disks only which won't work. Of course. Is there a way to force-assemble an array with an arbitrary sequence of devices? Also, according to mdadm -E on the disks, the array wasn't actually grown to 5 devices: > Raid Devices : 4 Also, the size of the array: > Array Size : 5860558848 (2794.53 GiB 3000.61 GB) suggests the array is actually still a 4-disk one, so 3 working disks should constitute a degraded array. I understand that while the array was being reshaped, data was moved onto the 4th disk, too, so all the 4 devices will be needed to recover. Why was sdf marked as a spare and kicked out, I can't understand. > I faced interruption during a grow process before, and re-issuing the > grow command worked for me, but I suggest you run all the tests > possible now and wait till more people reply to your email. sdh reported one unreadable sector. That suggests everything else should be possible to recover, right? BTW, bad sectors, as bad as they are, usually don't mean the disk failing completely, wouldn't it be better for the array to shut down with a huge warning that it will fail if this disk gets more errors instead of breaking the array if there're no spares available? Also, does the mentioned smartd polling/testing reduce the probability of the array failing? If the sectors are marked bad, the controller will simply redirect requests to them, right? So the disk wouldn't be marked as failed, but smartd will report problems early? -- Regards Michał (Saviq) Sawicz
Attachment:
signature.asc
Description: To jest =?UTF-8?Q?cz=C4=99=C5=9B=C4=87?= =?UTF-8?Q?_wiadomo=C5=9Bci?= podpisana cyfrowo
