Re: BUG:write data to degrade raid5

[Joachim Otahal]

Kristleifur Daðason schrieb:
> On Fri, Mar 19, 2010 at 6:20 PM, Joachim Otahal <Jou@xxxxxxx 
> <mailto:Jou@xxxxxxx>> wrote:
>
>     jin zhencheng schrieb:
>
>         hi;
>
>         i use kernel is 2.6.26.2
>
>         what i do as follow:
>
>         1, I create a raid5:
>         mdadm -C /dev/md5 -l 5 -n 4 /dev/sda /dev/sdb /dev/sdc  /dev/sdd
>         --metadata=1.0 --assume-clean
>
>         2, dd if=/dev/zero of=/dev/md5 bs=1M&
>
>         write data to this raid5
>
>         3, mdadm --manage /dev/md5 -f /dev/sda
>
>         4 mdadm --manage  /dev/md5 -f /dev/sdb
>
>         if i faild 2 disks ,then the OS kernel display OOP error and
>         kernel down
>
>         do somebody know why ?
>
>         Is MD/RAID5 bug ?
>
>
>     RAID5 can only tolerate ONE drive to fail of ALL members. If you
>     want to be able to fail two drives you will have to use RAID6 or
>     RAID5 with one hot-spare (and give it time to rebuild before
>     failing the second drive).
>     PLEASE read the documentation on raid levels, like on wikipedia.
>
>
> That is true,
>
> but should we get a kernel oops and crash if two RAID5 drives are 
> failed? (THAT part looks like a bug!)
>
> Jin, can you try a newer kernel, and a newer mdadm?
>
> -- Kristleifur
You are probably right.
My kernel version is "Debian 2.6.26-21lenny4", and I had no oopses 
during my hot-plug testing one the hardware I use md on. I think it may 
be the driver for his chips.

Jin:

Did you really use the whole drives for testing or loopback files or 
partitions on the drives? I never did my hot-plug testings with whole 
drives being in an array, only with partitions.

Joachim Otahal

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/