growing md2, do I need three reboots?

Hello,

my /dev/md2 uses superblock 1.0, which is stored at the end of
device. Therefore I suppose that this approach of growing it isn't
going to work:

#Alter the partition tables, to make /dev/sd[abc]2 have new size:
fdisk /dev/sda
fdisk /dev/sdb
fdisk /dev/sdc

#reboot (make kernel read new partition table)

#then, grow the raid1 /dev/md1
mdadm --grow /dev/md2 --size=max

#and finaly, grow the ext2 (or ext3) fs on /dev/md1
resize2fs /dev/md1

Because in this way, after rebooting, /dev/md2 won't be found - the
superblock won't be in the correct place.

So, I would need to remove each of them from the array, resize
partition, then add it back. Thus needing three reboots, since I can
remove only one device at a time (the same HDDs have also partitions
belonging to root raid1 array), which must be always up & running.

I can afford reboots, no problem here, but isn't there some simpler way?


below is my raid layout, I need to grow md2 by few spare gigabytes
left at the end of /dev/sd[abc].

kernel 2.6.29 (impossible to upgrade at the moment).

Personalities : [raid0] [raid1] [raid10]
md2 : active raid10 sda2[0] sdc2[2] sdb2[1]
185381376 blocks super 1.0 512K chunks 2 far-copies [3/3] [UUU]
bitmap: 1/6 pages [4KB], 16384KB chunk

md1 : active raid1 sdc1[2](W) sdb1[3](W)
9767416 blocks super 1.0 [2/2] [UU]
bitmap: 1/150 pages [4KB], 32KB chunk

md0 : active raid1 sde1[0] sdd1[2] sda1[1]
9767424 blocks [3/3] [UUU]
bitmap: 1/150 pages [4KB], 32KB chunk

unused devices:
atak:/home/janek# mdadm -D /dev/md2
/dev/md2:
Version : 1.0
Creation Time : Thu Sep 2 11:47:39 2010
Raid Level : raid10
Array Size : 185381376 (176.79 GiB 189.83 GB)
Used Dev Size : 123587584 (117.86 GiB 126.55 GB)
Raid Devices : 3
Total Devices : 3
Persistence : Superblock is persistent

Intent Bitmap : Internal

Update Time : Thu Dec 2 16:41:02 2010
State : active
Active Devices : 3
Working Devices : 3
Failed Devices : 0
Spare Devices : 0

Layout : far=2
Chunk Size : 512K

Name : atak:2 (local to host atak)
UUID : f2a75dbe:5ac91a1f:c09da3c0:f6f69c9c
Events : 28

Number Major Minor RaidDevice State
0 8 2 0 active sync /dev/sda2
1 8 18 1 active sync /dev/sdb2
2 8 34 2 active sync /dev/sdc2

best regards
--
Janek Kozicki http://janek.kozicki.pl/ |
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: growing md2, do I need three reboots?

On Thu, 2 Dec 2010 16:52:03 +0100 Janek Kozicki wrote:

> Hello,
>
> my /dev/md2 uses superblock 1.0, which is stored at the end of
> device. Therefore I suppose that this approach of growing it isn't
> going to work:
>
> #Alter the partition tables, to make /dev/sd[abc]2 have new size:
> fdisk /dev/sda
> fdisk /dev/sdb
> fdisk /dev/sdc
>
> #reboot (make kernel read new partition table)
>
> #then, grow the raid1 /dev/md1
> mdadm --grow /dev/md2 --size=max
>
> #and finaly, grow the ext2 (or ext3) fs on /dev/md1
> resize2fs /dev/md1
>
> Because in this way, after rebooting, /dev/md2 won't be found - the
> superblock won't be in the correct place.
>
> So, I would need to remove each of them from the array, resize
> partition, then add it back. Thus needing three reboots, since I can
> remove only one device at a time (the same HDDs have also partitions
> belonging to root raid1 array), which must be always up & running.
>
> I can afford reboots, no problem here, but isn't there some simpler way?

Yes, there is a simpler way, but no: it isn't going to work anyway.

You cannot 'grow' a RAID10 array at all - sorry. It is sufficiently complex
that it needs quite a bit of time to design, code, and test. And I haven't
had that time yet.

But if you could resize a RAID10 array, this is what I would do:

1/ For each devices (sda, sdb, sdc)
- fail and remove each partition from the respective array.
- run 'kpartx -a /dev/sdX'. This will create partitions in
/dev/mapper/ with the same names.
- --re-add these partitions to the arrays. The presence of a
write-intent-bitmap will mean that resync is almost instant.

2/ Use fdisk to change the partition tables.

3/ run 'kpartx -a /dev/sdX' again on each device. This will change the
partitions even while they are active.

4/ For the partitions which have changed size, find the matching
/dev/md2/md/dev-dm0X/size
and
echo 0 > /dev/md2/md/dev-dm-X/size

This will cause md to relocate the metadata to the new end of the device.
Not that these partitions (created by kpartx) are device-mapper partitions
so have names like 'dm-0' and 'dm-1'.

5/ mdadm -G /dev/md2 --size max
This bit unfortunately won't work.


NeilBrown



>
>
> below is my raid layout, I need to grow md2 by few spare gigabytes
> left at the end of /dev/sd[abc].
>
> kernel 2.6.29 (impossible to upgrade at the moment).
>
> Personalities : [raid0] [raid1] [raid10]
> md2 : active raid10 sda2[0] sdc2[2] sdb2[1]
> 185381376 blocks super 1.0 512K chunks 2 far-copies [3/3] [UUU]
> bitmap: 1/6 pages [4KB], 16384KB chunk
>
> md1 : active raid1 sdc1[2](W) sdb1[3](W)
> 9767416 blocks super 1.0 [2/2] [UU]
> bitmap: 1/150 pages [4KB], 32KB chunk
>
> md0 : active raid1 sde1[0] sdd1[2] sda1[1]
> 9767424 blocks [3/3] [UUU]
> bitmap: 1/150 pages [4KB], 32KB chunk
>
> unused devices:
> atak:/home/janek# mdadm -D /dev/md2
> /dev/md2:
> Version : 1.0
> Creation Time : Thu Sep 2 11:47:39 2010
> Raid Level : raid10
> Array Size : 185381376 (176.79 GiB 189.83 GB)
> Used Dev Size : 123587584 (117.86 GiB 126.55 GB)
> Raid Devices : 3
> Total Devices : 3
> Persistence : Superblock is persistent
>
> Intent Bitmap : Internal
>
> Update Time : Thu Dec 2 16:41:02 2010
> State : active
> Active Devices : 3
> Working Devices : 3
> Failed Devices : 0
> Spare Devices : 0
>
> Layout : far=2
> Chunk Size : 512K
>
> Name : atak:2 (local to host atak)
> UUID : f2a75dbe:5ac91a1f:c09da3c0:f6f69c9c
> Events : 28
>
> Number Major Minor RaidDevice State
> 0 8 2 0 active sync /dev/sda2
> 1 8 18 1 active sync /dev/sdb2
> 2 8 34 2 active sync /dev/sdc2
>
> best regards

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: growing md2, do I need three reboots?

Neil Brown said: (by the date of Fri, 3 Dec 2010 12:25:47 +1100)

> > I can afford reboots, no problem here, but isn't there some simpler way?
>
> Yes, there is a simpler way, but no: it isn't going to work anyway.

Great! Thank you for your reply, I must remember this in case when
I'll want to grow a non-raid10 array :)

And now: LVM to the rescue!

best regards
Janek Kozicki



> You cannot 'grow' a RAID10 array at all - sorry. It is sufficiently complex
> that it needs quite a bit of time to design, code, and test. And I haven't
> had that time yet.
>
> But if you could resize a RAID10 array, this is what I would do:
>
> 1/ For each devices (sda, sdb, sdc)
> - fail and remove each partition from the respective array.
> - run 'kpartx -a /dev/sdX'. This will create partitions in
> /dev/mapper/ with the same names.
> - --re-add these partitions to the arrays. The presence of a
> write-intent-bitmap will mean that resync is almost instant.
>
> 2/ Use fdisk to change the partition tables.
>
> 3/ run 'kpartx -a /dev/sdX' again on each device. This will change the
> partitions even while they are active.
>
> 4/ For the partitions which have changed size, find the matching
> /dev/md2/md/dev-dm0X/size
> and
> echo 0 > /dev/md2/md/dev-dm-X/size
>
> This will cause md to relocate the metadata to the new end of the device.
> Not that these partitions (created by kpartx) are device-mapper partitions
> so have names like 'dm-0' and 'dm-1'.
>
> 5/ mdadm -G /dev/md2 --size max
> This bit unfortunately won't work.
>
>
> NeilBrown
>
>
>
> >
> >
> > below is my raid layout, I need to grow md2 by few spare gigabytes
> > left at the end of /dev/sd[abc].
> >
> > kernel 2.6.29 (impossible to upgrade at the moment).
> >
> > Personalities : [raid0] [raid1] [raid10]
> > md2 : active raid10 sda2[0] sdc2[2] sdb2[1]
> > 185381376 blocks super 1.0 512K chunks 2 far-copies [3/3] [UUU]
> > bitmap: 1/6 pages [4KB], 16384KB chunk
> >
> > md1 : active raid1 sdc1[2](W) sdb1[3](W)
> > 9767416 blocks super 1.0 [2/2] [UU]
> > bitmap: 1/150 pages [4KB], 32KB chunk
> >
> > md0 : active raid1 sde1[0] sdd1[2] sda1[1]
> > 9767424 blocks [3/3] [UUU]
> > bitmap: 1/150 pages [4KB], 32KB chunk
> >
> > unused devices:
> > atak:/home/janek# mdadm -D /dev/md2
> > /dev/md2:
> > Version : 1.0
> > Creation Time : Thu Sep 2 11:47:39 2010
> > Raid Level : raid10
> > Array Size : 185381376 (176.79 GiB 189.83 GB)
> > Used Dev Size : 123587584 (117.86 GiB 126.55 GB)
> > Raid Devices : 3
> > Total Devices : 3
> > Persistence : Superblock is persistent
> >
> > Intent Bitmap : Internal
> >
> > Update Time : Thu Dec 2 16:41:02 2010
> > State : active
> > Active Devices : 3
> > Working Devices : 3
> > Failed Devices : 0
> > Spare Devices : 0
> >
> > Layout : far=2
> > Chunk Size : 512K
> >
> > Name : atak:2 (local to host atak)
> > UUID : f2a75dbe:5ac91a1f:c09da3c0:f6f69c9c
> > Events : 28
> >
> > Number Major Minor RaidDevice State
> > 0 8 2 0 active sync /dev/sda2
> > 1 8 18 1 active sync /dev/sdb2
> > 2 8 34 2 active sync /dev/sdc2
> >
> > best regards
>
>


--
Janek Kozicki http://janek.kozicki.pl/ |
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html