[arch-general] dmraid - Array failed - installed new pair of drives - any shortcuts for reinstall on new array?

David C. Rankin drankinatty at suddenlinkmail.com
Sat Jul 23 13:51:52 EDT 2011


On 07/22/2011 07:46 PM, David C. Rankin wrote:
>    Is there any shortcut to install Arch since I have the old install currently
> running? Or is it just easier to stick the install CD in and go through the base
> install on the new disks and then copy the old /var/cache/pacman/pkg to the new
> arrays to complete the package install with local packages?
>
>    I'm fairly sure the dmraid designation with the old drive will take care of
> itself when I do the install on the new array, but I'm not certain.

Lesson learned:

Warning: NEVER delete a partition in cfdisk to create 2 partitions with dmraid 
after Manually configure block devices, filesystems and mountpoints have been 
set. (really screws with dmraid metadata and existing partitions are worthless) 
Solution: delete the array from the bios and re-create to force creation under a 
new /dev/mapper ID, reinstall/repartition.
(added to: 
https://wiki.archlinux.org/index.php/Installing_with_Fake_RAID#Mounting_the_filesystem)

[12:16 nirvana:/home/david] # df -h
Filesystem                      Size  Used Avail Use% Mounted on
udev                             10M     0   10M   0% /dev
run                              10M  172K  9.9M   2% /run
/dev/mapper/nvidia_ddddhhfhp5    23G   13G  9.4G  57% /
shm                             1.9G     0  1.9G   0% /dev/shm
/dev/mapper/nvidia_ddddhhfhp10  608G  380G  198G  66% /home
/dev/mapper/nvidia_ddddhhfhp7   122M   34M   82M  30% /boot
/dev/mapper/nvidia_ddddhhfhp8    23G  6.0G   16G  28% /var
/dev/mapper/nvidia_ddddhhfhp9    33G  8.1G   23G  27% /srv
/dev/mapper/nvidia_edfgdeacp5    28G  1.2G   25G   5% /mnt/nv2
/dev/mapper/nvidia_edfgdeacp7   183M   35M  139M  21% /mnt/nv2/boot
/dev/mapper/nvidia_edfgdeacp8   851G  200M  808G   1% /mnt/nv2/home
/dev/mapper/nvidia_edfgdeacp9    37G  176M   35G   1% /mnt/nv2/srv

New array is nvidia_edfgdeac under temp mount point of /mnt/nv2.

Is there any downside to just copying the system from old drive to new array 
with (https://bbs.archlinux.org/viewtopic.php?id=77419):

1. Boot livecd
2. mount /dev/sda1 /mnt/old
3. mount /dev/sdb1 /mnt/new
4. cp -rav /mnt/old/* /mnt/new
5. umount /mnt/old
6. mount -o bind /dev /mnt/new/dev
7. mount -t proc none /mnt/new/proc
8. chroot /mnt/new /bin/bash
---1. grub-install /dev/sdb
---2. exit
9. Exit and reboot

Or is it better to just to selectively move data (i.e. mysqldump 
--all-databases) and then read it back in? Does it matter? Obviously, it would 
be cleaner just to reconfigure/reload all data to avoid copying stuff you don't 
need, but from a time standpoint the copy method is very attractive. Anybody 
have any thoughts either way?

-- 
David C. Rankin, J.D.,P.E.


More information about the arch-general mailing list