Robert Elz <kre%munnari.OZ.AU@localhost> writes: > Date: Mon, 29 Feb 2016 08:38:31 +0000 > From: Emmanuel Dreyfus <manu%netbsd.org@localhost> > Message-ID: <20160229083831.GF7353%homeworld.NetBSD.org@localhost> > > | I have trouble to find an explanation of how both events occurend. > > One way that raidframe might be involved, is that it writes whole > stripes at a time, so it can easily be writing blocks that you never > expected it to write to. Depending upon your stripe sizes, for example, > it might write to the boot block while updating the filesystem superblock. > > If you had bad ram (or dma, or something) one or both of the writes > might go wrong, considering that raid1 has to update both copies on a write. FWIW, I had 'raidframe corruption' once where the two RAID1 disks ended up different. It was in fact bad RAM. One thing you can do is to compare the two halves. You can mount them read-only with shadow disklabels. and you can umount the raidset, and then run fsck -n on each half after you run fsck on the set.
Attachment:
signature.asc
Description: PGP signature