cuak

David Véjar

Soporte y Redes

Ferretería Santiago

Soluciones de Abastecimiento

Lira 919 Santiago, Chile

Fono: (56 2) 731 3824

www.ferreteriasantiago.cl

-----Mensaje original----- De: centos-bounces@centos.org [mailto:centos-bounces@centos.org] En nombre de John R Pierce Enviado el: jueves, 12 de agosto de 2010 18:22 Para: CentOS mailing list Asunto: Re: [CentOS] Problem resizing partition of nfs volume

On 08/12/10 2:56 PM, Dan Yamins wrote:

Hi:

I have an NFS volume that I'm trying to resize a partition on.

Something about the fdisk process is corrupting something on the drive

Before running fdisk, I can mount the volume find:

$ mount /dev/sdo1 /home

... and the volume is mounted fine.

And, $ e2fsck -f /dev/sdo1 /dev/sdo1: clean, ...

But then I run fdisk to rewrite the partition table of this drive, to expand the /dev/sdo1 partition w/o losing data:

$ fdisk /dev/sdo # Type 'd' to delete the primary partition # Type 'n' for new partition # Type 'p' for primary # Type '1' for 1st # Type Enter for 1st cylinder # Type Enter for last cylinder (full disk) # Type 'w' to finish

Calling ioctl() ... Syncing disks ...

But then something has gone wrong.

$ e2fsck -f /dev/sdo1 e2fsck 1.39 (29-May-2006) Couldn't find ext2 superblock, trying backup blocks... e2fsck: Bad magic number in super-block while trying to open /dev/sdm1 ...

And I can't mount the volume any more: $ mount /dev/sdo1 /home mount: you must specify the filesystem type

What am I doing wrong? Am I missing a step of the process?

yeah, you deleted your file system, then created a new empty partition.

instead, you should have used parted(8) or similar to expand the partition, then used resize2fs(8) to expand the file system on this partition to its new size.

me, I'd strongly suggest doing a full backup of the partition first with dump(8)

_______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos

On Thu, Aug 12, 2010 at 6:21 PM, John R Pierce pierce@hogranch.com wrote:

On 08/12/10 2:56 PM, Dan Yamins wrote:

...

Hi:

I have an NFS volume that I'm trying to resize a partition on.

Something about the fdisk process is corrupting something on the drive

Before running fdisk, I can mount the volume find:

$ mount /dev/sdo1 /home

... and the volume is mounted fine.

And, $ e2fsck -f /dev/sdo1 /dev/sdo1: clean, ...

But then I run fdisk to rewrite the partition table of this drive, to expand the /dev/sdo1 partition w/o losing data:

$ fdisk /dev/sdo # Type 'd' to delete the primary partition # Type 'n' for new partition # Type 'p' for primary # Type '1' for 1st # Type Enter for 1st cylinder # Type Enter for last cylinder (full disk) # Type 'w' to finish

Calling ioctl() ... Syncing disks ...

But then something has gone wrong.

$ e2fsck -f /dev/sdo1 e2fsck 1.39 (29-May-2006) Couldn't find ext2 superblock, trying backup blocks... e2fsck: Bad magic number in super-block while trying to open /dev/sdm1

...

...

And I can't mount the volume any more: $ mount /dev/sdo1 /home mount: you must specify the filesystem type

What am I doing wrong? Am I missing a step of the process?

yeah, you deleted your file system, then created a new empty partition.

Luckily I'm doing this via snapshots of an amazon EBS volume ... so this whole thing was done on throw-away-able backup volume, which can be recreated from the original volume at any time. I've just done this, and I've attached the newly-created volume to /dev/sdo.

Now I'm trying to use parted.

[root@domU-12-31-39-0E-B2-61 ~]# parted /dev/sdo Error: Error initialising SCSI device /dev/sdo - Invalid argument

I _can_ do it on the volume path itself:

[root@domU-12-31-39-0E-B2-61 ~]# parted /dev/sdo1 Using /dev/sdo1 Welcome to GNU Parted! Type 'help' to view a list of commands. (parted)

.. but this seems wrong, since I'm trying to edit the partition table not of the partition /dev/sdo1 but of the drive itself. ... Or perhaps I should this? (If so, could you give more explicit instructions?)

Perhaps because this is a network attached drive, something virtual that amazon provides. However, the reason I've written to the centOS list is because the same procedure has worked fine with ubuntu images ... so it didnt' seem like an amazon problem

Thanks! Dan