grow raid 6: different errors with mdadm 3.0.3 and 3.1

grow raid 6: different errors with mdadm 3.0.3 and 3.1

[Andrew Burgess]

mdadm 3.0.3:

# mdadm -v /dev/md1 --grow --raid-devices=15
mdadm: Need to backup 19968K of critical section..
mdadm: /dev/md1: failed to save critical region

mdadm 3.1:

sudo ./mdadm -G /dev/md1 --raid-devices=15
mdadm: this change will reduce the size of the array.
        use --grow --array-size first to truncate array.
        e.g. mdadm --grow /dev/md1 --array-size 928142080

Hmmm. Can I just echo something into /proc to make this happen?
(assuming these are both mdadm bugs)

Or should I try another mdadm? I googled and a problem like #1
occured to someone using 3.0 and they worked around by using
2.6.7. But I'm starting to get nervous that mdadm may really be
trying to stop me from doing something dumb...

New fc12 upgrade, kernel 2.6.31.5-127.fc12.x86_64

Thanks for any help

------------------------------

sudo ./mdadm -D /dev/md1
/dev/md1:
         Version : 1.02
   Creation Time : Tue Sep 11 13:49:50 2007
      Raid Level : raid6
      Array Size : 8785916928 (8378.90 GiB 8996.78 GB)
   Used Dev Size : 732159744 (698.24 GiB 749.73 GB)
    Raid Devices : 14
   Total Devices : 15
     Persistence : Superblock is persistent

     Update Time : Tue Nov 24 10:05:31 2009
           State : clean
  Active Devices : 14
Working Devices : 15
  Failed Devices : 0
   Spare Devices : 1

          Layout : left-symmetric
      Chunk Size : 128K

            Name : 5
            UUID : 5211e02e:19f8718e:8feeb699:03a4879f
          Events : 9329236

     Number   Major   Minor   RaidDevice State
        0       8       17        0      active sync   /dev/sdb1
       10       8      113        1      active sync   /dev/sdh1
        2       8       65        2      active sync   /dev/sde1
        8       8      161        3      active sync   /dev/sdk1
        4       8       81        4      active sync   /dev/sdf1
       13       8      129        5      active sync   /dev/sdi1
        9       8       97        6      active sync   /dev/sdg1
       11       8       33        7      active sync   /dev/sdc1
       12       8      177        8      active sync   /dev/sdl1
        7       8      145        9      active sync   /dev/sdj1
       15      65       49       10      active sync   /dev/sdt1
       14       8       49       11      active sync   /dev/sdd1
       17       9        4       12      active sync   /dev/md4
       16       9        3       13      active sync   /dev/md3

       18      65        1        -      spare   /dev/sdq1

Re: grow raid 6: different errors with mdadm 3.0.3 and 3.1

[Neil Brown]

On Tue, 24 Nov 2009 10:21:39 -0800
Andrew Burgess <aab@cichlid.com> wrote:

> 
> mdadm 3.0.3:
> 
> # mdadm -v /dev/md1 --grow --raid-devices=15
> mdadm: Need to backup 19968K of critical section..
> mdadm: /dev/md1: failed to save critical region
> 
> mdadm 3.1:
> 
> sudo ./mdadm -G /dev/md1 --raid-devices=15
> mdadm: this change will reduce the size of the array.
>         use --grow --array-size first to truncate array.
>         e.g. mdadm --grow /dev/md1 --array-size 928142080
> 
> Hmmm. Can I just echo something into /proc to make this happen?
> (assuming these are both mdadm bugs)
> 
> Or should I try another mdadm? I googled and a problem like #1
> occured to someone using 3.0 and they worked around by using
> 2.6.7. But I'm starting to get nervous that mdadm may really be
> trying to stop me from doing something dumb...

Thanks for the report.
There appears to be a big in mdadm-3.1.1 such that the new 
array size gets truncated to 32bits in this calculation so
it looks like it is getting smaller.

The following patch will fix it, and can also be collected
from my git tree (git://neil.brown.name/md).

Thanks,
NeilBrown

From 2ed4f75388f99968be58097941a9704f6e42d701 Mon Sep 17 00:00:00 2001
From: NeilBrown <neilb@suse.de>
Date: Thu, 26 Nov 2009 14:19:26 +1100
Subject: [PATCH] Grow: avoid truncation error when checking size of array.

array.size is only 32bit so it is not safe to multiply it
up before casting to (long long).
Actually, we shouldn't be using array.size here at all, but that
will get fixed in a subsequent patch.

Reported-by: Andrew Burgess <aab@cichlid.com>
Signed-off-by: NeilBrown <neilb@suse.de>
---
 Grow.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/Grow.c b/Grow.c
index 7764bdb..a654d4e 100644
--- a/Grow.c
+++ b/Grow.c
@@ -891,7 +891,7 @@ int Grow_reshape(char *devname, int fd, int quiet, char *backup_file,
 		}
 
 		/* Check that we can hold all the data */
-		size = ndata * array.size;
+		size = ndata * (long long)array.size;
 		get_dev_size(fd, NULL, &array_size);
 		if (size < (array_size/1024)) {
 			fprintf(stderr, Name ": this change will reduce the size of the array.\n"
-- 
1.6.4.3

Re: grow raid 6: different errors with mdadm 3.0.3 and 3.1

[Andrew Burgess]

On 11/25/2009 07:22:27 PM, Neil Brown wrote:
> On Tue, 24 Nov 2009 10:21:39 -0800 Andrew Burgess <aab@cichlid.com>  
> wrote:

> > mdadm 3.1:
> >
> > sudo ./mdadm -G /dev/md1 --raid-devices=15
> > mdadm: this change will reduce the size of the array.
> >         use --grow --array-size first to truncate array.
> >         e.g. mdadm --grow /dev/md1 --array-size 928142080


> Thanks for the report.

Thanks for the fix. I grabbed the git head version and the
resize is grinding away now.

Re: grow raid 6: different errors with mdadm 3.0.3 and 3.1

[Andrew Burgess]

On 11/24/2009 10:21:39 AM, Andrew Burgess wrote:

> mdadm 3.1:

oops, it seems to be 3.1.1 not 3.1
from  
http://neil.brown.name/git?p=mdadm;a=snapshot;h=40bc78f5cd292d90917cb0a8c177498a516494c3;sf=tgz