Re: [gentoo-user] Process to resize ext3 file system at the command line?

[Michael Mol]

On Thu, Nov 17, 2011 at 1:29 PM, Mark Knecht <markknecht@×××××.com> wrote:
> On Thu, Nov 17, 2011 at 9:25 AM, Neil Bothwick <neil@××××××××××.uk> wrote:
>> On Thu, 17 Nov 2011 09:01:46 -0800, Mark Knecht wrote:
>>
>>>    I'm pretty sure I've got the command set right to do the RAID-1 to
>>> RAID-5 conversion, but once it's done I believe the file system itself
>>> will still be 250GB so I'll need to resize the file system. In the
>>> past I've done this with gparted, which seems to work fine, but this
>>> time I was considering doing it at the command line. Does anyone know
>>> of a good web site that goes through how to do that? I've browsed
>>> around and found different pages that talk about it but my reading
>>> looks like they all have minor differences which leaves me a bit
>>> worried.
>>
>> Using cfdisk or fdisk, delete the partition and recreate it, USING THE
>> SAME START BLOCK at a larger size.
>>
>> Then "resize2fs /dev/sdwhatever" will resize the filesystem to fill the
>> partition.
>>
>>
>>
>> --
>> Neil Bothwick
>
> Really? Delete the partition? Sounds scary! (But actually makes sense.
> The data is still there.)
>
> I'm not sure how this works in the case of a RAID though. Here's the
> current partition table for sda where sda6, sdb6 & sdc6 are part of
> the RAID-1::
>
> c2stable ~ # fdisk -l /dev/sda
>
> Disk /dev/sda: 500.1 GB, 500107862016 bytes
> 255 heads, 63 sectors/track, 60801 cylinders, total 976773168 sectors
> Units = sectors of 1 * 512 = 512 bytes
> Sector size (logical/physical): 512 bytes / 512 bytes
> I/O size (minimum/optimal): 512 bytes / 512 bytes
> Disk identifier: 0x8b45be24
>
>   Device Boot      Start         End      Blocks   Id  System
> /dev/sda1   *          63      112454       56196   83  Linux
> /dev/sda2          112455     8514449     4200997+  82  Linux swap / Solaris
> /dev/sda3         8594775   113467094    52436160   fd  Linux raid autodetect
> /dev/sda4       113467095   976768064   431650485    5  Extended
> /dev/sda5       113467158   218339414    52436128+  fd  Linux raid autodetect
> /dev/sda6       481933935   976768064   247417065   83  Linux
> /dev/sda7       218339478   481933871   131797197   fd  Linux raid autodetect
>
> Partition table entries are not in disk order
> c2stable ~ #
>
> It's not that I want to change the partition size of the 3 pieces of
> the RAID-1, it's that after I convert the RAID-1 to RAID-5 I want it
> to be 500GB.
>
>
> I asked some questions on the Linux RAID list and putting together
> info from a couple of people here's how I'm thinking I proceed with
> the conversion:
>
> 1) First, fail one disk and clean it up for later:
>
> umount /dev/md6
> mdadm --stop /dev/md6
> mdadm /dev/md6 --fail /dev/sdc6 --remove /dev/sdc6
> mdadm --zero-superblock /dev/sdc6
>
> At this point the RAID-1 is still 3-drives but one is marked 'failed'.
> The failed drive is at this point like a new drive as it has no
> superblock. (I think...)
>
> 2) Now I convert the 3-drive RAID1 to a 2-drive RAID-1:
>
> mdadm --grow /dev/md6 --raid-devices=2
>
> 3) Create a 2-drive RAID-5:
>
> mdadm has an 'instantaneous' conversion of RAID-1 to RAID-5 for the
> 2-drive case because parity of a single drive is just the data itself.
> /dev/sdb6 is now 'parity' instead of 'data'.
>
> mdadm /dev/md6 --grow --level=5
>
> 4) Add a 3rd drive to the RAID-5:
>
> mdadm /dev/md6 --add /dev/sdc6
> mdadm /dev/md6 --grow --raid-devices=3
>
>
>
> At this point I was told:
>
> "Now, resize your filesystem to use the additional space."
>
> So, if at this point the end-block of sda6 isn't 976768064 but, let's
> say, 700000000 because mdadm set it to something new, then using your
> suggestion I guess I'd set it back to 976768064? I'm not comfortable
> however that if I do that that whatever is out there beyond 700000000
> is really formatted as ext3 and 'empty' as I don't know what the mdadm
> conversion has done to it.

Your resize would be applied not to /dev/sd?, but to /dev/md?. You
don't need to worry about what that means on /dev/sd*; the filesystem
you want to poke is on /dev/md*.

file -s /dev/sd* /dev/md*

-- 
:wq