On January 14, [EMAIL PROTECTED] wrote:
> How I have to start a RAID-10 array after disk failure?
>
> My RAID-10 is a RAID-1 array /dev/md05 overtwo RAID-0 arrays /dev/md50
> and /dev/md51. The two RAID-0 arrays are over partitions /dev/hd[ik]5
> and /dev/hd[eg]5.
>
> The array /dev/md05 is running fine. Now I've tested an failure
> situation: I shut down, disconnect the hdk disk and reboot. /dev/md51
> would autostart by kernel. /proc/mdstat says:
>
> Personalities : [linear] [raid0] [raid1] [raid5]
> read_ahead 1024 sectors
> md51 : active raid0 hdg5[1] hde5[0]
> 7369216 blocks 4k chunks
>
> unused devices: <none>
>
> Now I try to start /dev/md05:
>
> raidstart /dev/md05
>
> and get the error:
>
> could not import md50!
> autostart md50 failed!
> /dev/md05 invalid argument
>
> Why?
Because "raidstart" is overly simplistic. It takes the first device
mentioned in /etc/raidtab for the array in question and gives it to
the kernel saying
"go read the raid superblock of this device and find the other
devices in the array from the superblock and make it all work".
Ofcourse, if that first device isn't working, you loose.
There is a better interface to the kernel in 2.4 to allow raidstart
(or a similar program) to say:
Make me a raid array using this device, and this device, and this
device - check that the super blocks are consistant and do the right
thing.
but there isn't a program that makes use of this interface
yet... maybe one day.
>
> I can start /dev/md05 after editing /etc/raidtab, set /dev/md51
> as failure-disk and call
>
> mkraid -f /dev/md05
>
> Now /dev/md05 is running. If I reboot, again raidstart failed an I must
> use mkraid.
>
>
> Is this behavior intended? And if, is this documented? And if, where?
No (accidental), no, n/a.
>
> Peter
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to [EMAIL PROTECTED]
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to [EMAIL PROTECTED]
