Hi,
thanks to Jakobs really helpful HOWTO I managed to set up
a well-working RAID1 based on the 990309-edition, inklusive
RAID1 of the boot partion. However, smaller problems remain:
- If you change the type of a partion to "FD", a active inode
will be kept for this partion after auto-detection, even if
does not contain a RAID superblock.
What happened, is that I wanted to build up a RAID partion.
I first changed the type from 83 to FD, and as an old-fashioned
man I rebooted the computer. When I tried to set up my RAID1
on that partion, mkraid aborted.
The failure of the fs_may_mount(newdev)-function seems to
be the reason. Even more, some other strange messages in
/var/log/messages seem to result from open, active inodes:
E.g.:
Mar 12 15:54:01 katheter kernel: VFS: inode busy on removed device 08:18
Mar 12 15:54:01 katheter kernel: VFS: inode busy on removed device 08:17
Mar 12 15:54:01 katheter kernel: VFS: inode busy on removed device 08:16
or:
Mar 12 15:54:51 katheter kernel: md: can not import sda2, has active inodes!
Mar 12 15:54:51 katheter kernel: md: error, md_import_device() returned -16
I suppose it will be easy to reproduce the bug: Empty a partion
e.g. by dd if=/dev/zero of=/dev/xxxx, change its partition type to
FD and reboot. Try to do anything with it, like mke2fs and a mount,
or, as intended, RAID.
2)
My *really* simple question is: Once I set one of two disks of a mirror
down (e.g. by this sequence: Shutdown, Removal of 1 disk, Start of
Linux, Shutdown, Insertion of the second disk, Start),
I won't get it up into the mirror again:
E.g:
Mar 12 21:27:09 katheter kernel: .
Mar 12 21:27:09 katheter kernel: considering sdb7 ...
Mar 12 21:27:09 katheter kernel: adding sdb7 ...
Mar 12 21:27:09 katheter kernel: adding sda7 ...
Mar 12 21:27:09 katheter kernel: created md1
Mar 12 21:27:09 katheter kernel: bind<sda7,1>
Mar 12 21:27:09 katheter kernel: bind<sdb7,2>
Mar 12 21:27:09 katheter kernel: running: <sdb7><sda7>
Mar 12 21:27:09 katheter kernel: now!
Mar 12 21:27:09 katheter kernel: sdb7's event counter: 00000037
Mar 12 21:27:09 katheter kernel: sda7's event counter: 00000031
Mar 12 21:27:09 katheter kernel: md: superblock update time inconsistency -- usi
ng the most recent one
Mar 12 21:27:09 katheter kernel: freshest: sdb7
Mar 12 21:27:09 katheter kernel: md: kicking non-fresh sda7 from array!
Mar 12 21:27:09 katheter kernel: unbind<sda7,1>
Mar 12 21:27:09 katheter kernel: export_rdev(sda7)
Mar 12 21:27:09 katheter kernel: md1: max total readahead window set to 128k
Mar 12 21:27:09 katheter kernel: md1: 1 data-disks, max readahead per data-disk:
128k
Mar 12 21:27:09 katheter kernel: raid1: device sdb7 operational as mirror 1
Mar 12 21:27:09 katheter kernel: raid1: md1, not all disks are operational -- tr
ying to recover array
Mar 12 21:27:09 katheter kernel: raid1: raid set md1 active with 1 out of 2 mirr
ors
Mar 12 21:27:09 katheter kernel: md: updating md1 RAID superblock on device
Mar 12 21:27:09 katheter kernel: sdb7 [events: 00000038](write) sdb7's sb offset
: 513984
Mar 12 21:27:09 katheter kernel: md: recovery thread got woken up ...
Mar 12 21:27:09 katheter kernel: md1: no spare disk to reconstruct array! -- con
tinuing in degraded mode
Mar 12 21:27:09 katheter kernel: md: recovery thread finished ...
How can I achieve what has formerly been done by chkraid? (Syncing the
mirror and taking it active?) I have no spare disk, I just want my
mirror back'n working :-)
Thanks a lot for any answers, and I hope my bug report will help,
and thanks for the work that has been done,
Georg v.Zezschwitz
