Re: [RFC] wrong behavior of re-adding a device

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Fri, Sep 21, 2018 at 5:51 AM Guoqing Jiang <gqjiang@xxxxxxxx> wrote:
>
>
>
> On 9/20/18 9:36 PM, Gi-Oh Kim wrote:
>
> On Wed, Sep 19, 2018 at 5:09 PM Gi-Oh Kim <gi-oh.kim@xxxxxxxxxxxxxxxx> wrote:
>
> Hi,
>
> I found a weird behavior of re-adding a device.
> I think it is a kernel bug.
> I would appreciate it if somebody can confirm if it is a bug or feature.
>
> I tested re-adding a device as following.
> 1. create md with ram0 and ram1
> 2. add ram2
> 3. grow raid-device number to 3
> 4. remove ram2
> 5. grow raid-device number to 2
> 6. add ram2
> 7. ram0 become faulty and ram2 become active
> 8. stop md
> 9. assemble md with ram0 and ram1 => fail because ram0 is faulty
>
> Hi,
>
> I checked the kernel function raid1_spare_active() in raid1.c and
> found out ram0 is set as faulty on purpose.
> If ram0 is set as fauly to replace it with ram2, i think it should be
> successful to assemble ram1 and ram2.
> But "mdadm -A /dev/md111 /dev/ram1 /dev/ram2" creates md111 with only ram2.
> I do not understand why it is necessary to set ram0 faulty.
>
>
> When ram2 is add back to array, seems array prefers to use the previous role
> which is recorded by saved_raid_disk.
>
> How can I re-add ram2 device as the spare device without setting ram0 faulty?
>
>
> I guess you can achieve the goal by remove the superblock of ram2 (before step 6).
> BTW, could you try the below change? I think it can fix the issue.

Yes, I think so.
But I think it would be better to set ram0 as spare, not faulty.

And another problem is "mdadm -A /dev/md111 /dev/ram1 /dev/ram2" does
not work well.
ram1 does not assembled to md111.
If ram0 should be faulty, I think assembling ram1 and ram2 should work.

>
> diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
> index 4e990246225e..1d54109071cc 100644
> --- a/drivers/md/raid1.c
> +++ b/drivers/md/raid1.c
> @@ -1734,6 +1734,7 @@ static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
>          */
>         if (rdev->saved_raid_disk >= 0 &&
>             rdev->saved_raid_disk >= first &&
> +           rdev->saved_raid_disk < conf->raid_disks &&
>             conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
>                 first = last = rdev->saved_raid_disk;
>
>
>
> Thanks,
> Guoqing



-- 
GIOH KIM
Linux Kernel Entwickler

ProfitBricks GmbH
Greifswalder Str. 207
D - 10405 Berlin

Tel:       +49 176 2697 8962
Fax:      +49 30 577 008 299
Email:    gi-oh.kim@xxxxxxxxxxxxxxxx
URL:      https://www.profitbricks.de

Sitz der Gesellschaft: Berlin
Registergericht: Amtsgericht Charlottenburg, HRB 125506 B
Geschäftsführer: Achim Weiss, Matthias Steinberg, Christoph Steffens




[Index of Archives]     [Linux RAID Wiki]     [ATA RAID]     [Linux SCSI Target Infrastructure]     [Linux Block]     [Linux IDE]     [Linux SCSI]     [Linux Hams]     [Device Mapper]     [Device Mapper Cryptographics]     [Kernel]     [Linux Admin]     [Linux Net]     [GFS]     [RPM]     [git]     [Yosemite Forum]


  Powered by Linux