[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Re: Re: mdadm serious does wrong partition changes



Hi Hermann,

On Mon, 2017-12-04 at 12:16 +0100, Hermann Lauer wrote:
> Hi Frans,
> 
> On Mon, Dec 04, 2017 at 11:53:07AM +0100, Frans van Berckel wrote:
> > And how people do a bootable device with mdraid, read /boot and
> > / mirrored, with Sun table. ;-)

> old:~# fdisk /dev/sda
> Disk /dev/sda (Sun disk label): 27 heads, 107 sectors, 24620
> cylinders
> Units = sectors of 1 * 512 bytes
> 
>    Device Flag    Start       End    Blocks   Id  System
> /dev/sda1             0    409600    204800    1  Boot
> /dev/sda2        410238  68420186  34004974   83  Linux native
> /dev/sda3             0  71132958  35566479    5  Whole disk
> /dev/sda4  u   68420187  71127180   1353496+  82  Linux swap

For what i found out, with Debian Sparc64 so far, talking about a Sun
table solution. Creating what you have now isn't possible any more.

> old:~# cat /proc/mdstat 

> Personalities : [raid1] 
> md1 : active raid1 sda1[0] sdb1[1]
>       204736 blocks [2/2] [UU]
>       
> md0 : active raid1 sda2[0] sdb2[1]
>       34004864 blocks [2/2] [UU]
> 
> sdb having *exactly* the same partioning and md1 running 0.90
> layout.

So md1 = sda1 (and sdb1) = boot, a less then 1GB with ext2 (or ext3).

1) Creating a md1 on sda1 (and sdb1) with a 1.2 layout will make sda2
and sda4 no longer accessible. And that's what Debian installer does.

2) Creating a md1 on sda1 (and sdb1) with a 0.90 layout works. It will
leaf sda2 and sda4 as it is.

But mdadm creates a md1p1 (and md1p3) partition on sda1 (and
sdb1) 'itselfs'. If you ignore p1 and creating a ext3 file system on
md1, it's gonna yell there's a partition table on it, do you wanna
overwrite it? It's the same counts for creating a ext3 on md1p1. So
there goes something wrong.

- To be clear, with the next example print, I didn't add md1 to my
mdadm.conf yet, so it's renamed by Debian to md127.

# lsblk

NAME          MAJ:MIN RM   SIZE RO TYPE  MOUNTPOINT
sda             8:0    0 136.7G  0 disk
├─sda1          8:1    0   957M  0 part  /boot
├─sda2          8:2    0 135.8G  0 part
│ └─md0         9:0    0 135.7G  0 raid1 /
└─sda3          8:3    0 136.7G  0 part
sdb             8:16   0 136.7G  0 disk
├─sdb1          8:17   0   957M  0 part
│ └─md127       9:127  0 956.9M  0 raid1
│   ├─md127p1 259:0    0 956.9M  0 md
│   └─md127p3 259:1    0 956.9M  0 md    /media/disk1
├─sdb2          8:18   0 135.8G  0 part
│ └─md0         9:0    0 135.7G  0 raid1 /
└─sdb3          8:19   0 136.7G  0 part

I think this isn't a bootable device after all. Because how is Sun OBP
(and Silo) gonna handle a sdb1 (and sda1) partition table, a
mdraid md127 device, with the md127p1 partition table on it while
booting? Or am i wrong?

And most old Sun OBP's aren't handling a GPT partition table for
booting? Or am i wrong? True, as Adrian ask for, i created GPT
partitions on sdc and sdd with a 1.2 layout, mounting /home and swap,
what's a working solution.

So summarizing, the problem is the boot partition.
And we wanna have bootable Sun server. ;-)

Thanks,


Frans van Berckel


Reply to: