On Tue, Oct 30, 2018 at 7:35 PM Lee Trager <lee.tra...@canonical.com> wrote:
>
> Every time I commission the PPC host we have in our CI I always get a
> GPT partitioning table using MAAS 2.4.2+.  Looking through the source
> code it appears GPT is always set when creating a new partitioning
> table[1] and when generating the preseed[2] due to the bios_boot_method
> being powernv. I'm not sure how msdos got set but I believe the
> partition table type is a red herring.

>
> The issue appears to be that Curtin is not wiping the PReP
> partition(/dev/sda1). If I chroot into the deployment environment and
> try to install GRUB with


Well, other than simple mode, custom storage mode has never
wiped a prep partition. while that's a bug, it doesn't explain why
we only see this issue now.


>
> grub-install --target=powerpc-ieee1275 /dev/sda1 --no-nvram
>
> it fails. If I wipe /dev/sda1 first GRUB installation succeeds. I can
> then retry deployment and deployment works fine. I suspect the issue has
> to do with a newer version of GRUB being installed. Previous versions of
> GRUB must of been similar enough that the PReP partition didn't have to
> be fully wiped before installing.

That makes more sense; if the layout changed, but I suspect that PReP format
isn't really changing.  Maybe it's attempting to copy in a new payload and
checking.  The check for clean prep partition is from 2013, so no new grub
code AFAICT.

>
> @Ryan, as you mentioned Curtin has a bug where it is not fully wiping
> the PReP partition which appears to be the root cause of this bug. It
> may be worth modifying Curtin to always wipe any partition with the
> 'prep' flag set regardless of what wipe is set to to ensure this doesn't
> happen in other environments.

That seems reasonable, I was adding a wipe after creating if the
mode wasn't superblock (as we already clear superblocks before
creating partitions to not trigger buried data like raid or bcache).

But just ensuring prep partition is wiped works as well.

I'll have a branch up soon.

I would like to understand what's different on the CI side as
the curtin code *hasnt* changed despite there being a bug
for not clearing it.

>
>
> [1] 
> https://git.launchpad.net/maas/tree/src/maasserver/models/partitiontable.py#n152
> [2] https://git.launchpad.net/maas/tree/src/maasserver/preseed_storage.py#n254
>
> ** Changed in: curtin
>        Status: Incomplete => Confirmed
>
> ** Changed in: grub2 (Ubuntu)
>        Status: New => Invalid
>
> ** Changed in: maas
>        Status: Incomplete => Invalid
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1800153
>
> Title:
>   [2.5] Failed to deploy ppc64el when partition table is GPT
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/curtin/+bug/1800153/+subscriptions

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1800153

Title:
  [2.5] Failed to deploy ppc64el when partition table is GPT

To manage notifications about this bug go to:
https://bugs.launchpad.net/curtin/+bug/1800153/+subscriptions

-- 
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to