Bug#706862: Database destroyed during upgrade from squeeze

[Ondřej Surý]

On Thu, May 9, 2013 at 4:18 AM, Ben Hutchings <ben at decadent.org.uk> wrote:
> On Mon, 2013-05-06 at 07:23 +0200, Ondřej Surý wrote:
>> I still miss the answer for:
>>
>> >> Could you check the permissions on /var/lib/cyrus and it's contents?
>
> This is from the full system backup that ran just before the upgrade:
>
[...]

I see nothing wrong in here.

>> And could it be that you ran out of free space in /tmp?
>
> It's a tmpfs which appears to to have a capacity of 2G (there is 4G of
> swap, barely used).  I don't know how much free space it had during the
> upgrade, of course.

That might be the reason. Your deliver databases very quite big, maybe
I just should stay on the same filesystem for the migration.

>> I have tested the migration script throughly, but there still might be
>> some corner cases unhandled.
>
> Well, note that migration was triggered by running the init script
> 'status' action (which is itself a bug - only 'start' should do that)
> while the upgraded package was in the unpacked state.  Are you sure that
> doesn't make a difference?

Could you please fill a separate bug report (I am sitting at the
Windows machine right now, which makes me pretty useless).

>> Well, it would help me to understand what has happened if I could test
>> with real data. So, yes, it would be nice to lay my hands on full
>> backup.
> [...]
>
> I'm afraid these databases seem to include private information that I am
> not prepared to disclose.

I understand.

> How about I try restoring the system backup on some other machine and
> re-run the upgrade with 'set -x' added to upgrade-db?

That would be great. I just need to know what has happened to fix it :(.

O.
--
Ondřej Surý <ondrej at sury.org>