Re: Need to move RAID1 with mounted partition

From: Leslie Rhorer <lesrhorer@att.net>
To: Roman Mamedov <rm@romanrm.net>
Cc: Linux RAID <linux-raid@vger.kernel.org>
Subject: Re: Need to move RAID1 with mounted partition
Date: Wed, 20 Apr 2022 04:21:15 -0500	[thread overview]
Message-ID: <03e77131-a306-f261-b9fd-47cff331ee37@att.net> (raw)
In-Reply-To: <20220420135533.39a32200@nvm>


On 4/20/2022 3:55 AM, Roman Mamedov wrote:
> On Wed, 20 Apr 2022 03:40:12 -0500
> Leslie Rhorer <lesrhorer@att.net> wrote:
> 
>> 	I have run into a little problem.  I know of a couple of ways to fix it
>> by shutting down the system and physically taking it apart, but for
>> various reasons I don't wish to take that route.  I want to be able to
>> re-arrange the system with it running.
>>
>> 	The latest version (bullseye) of Debian will not complete its upgrade
>> properly because my /boot file system is a little too small.  I have two
>> bootable drives with three partitions on them.  The first partition on
>> each drive is assembled into a RAID1 as /dev/md1 mounted as /boot.  Once
>> the system is booted, these can of course easily be umounted, the RAID1
>> stopped, and there is then no problem increasing the size of the
>> partitions if there were space to be had.  The third partition on each
>> drive is assigned as swap, and of course it was easy to resize those
>> partitions, leaving an additional 512MB between the second and third
>> partitions on each drive.  All I need to do is move the second partition
>> on each drive up by 512MB.
>>
>> 	The problem is the second partition on both drives is also assembled
>> into a RAID1 array on /dev/md2, formatted as ext4 and mounted as /.  Is
>> there a way I can move the RAID1 array up without shutting down the
>> system?  I don't need to resize the array, just move it.
> 
> You could fail one half of the RAID1, remove it, recreate the partition at the
> required offset, add the new partition into the array and let it rebuild. Then
> repeat with the other half.
> 
> However during that process you do not have the redundancy protection, so in
> case the remaining array drive fails or has a bad sector, it could become
> tricky to recover.
> 
> Maybe run a bad block scan, or "smartctl -t long" on the disks first. And of
> course have a backup.

	Hmm.  I hadn't thought of that.  Well, of course I thought of the 
backup.  I'm not insane.  The boot drives are SSDs, and they are not all 
that big.  The /dev/md2 array is only 88G, so there isn't much exposure 
to drive failure.  Rebuilding won't take long.  Thanks.