On ds., nov. 30 2019, Evilham wrote:

> On dv., nov. 29 2019, Evilham wrote:
>
>> On dv., nov. 29 2019, Mark Hindley wrote:
>>
>>> Evilham,
>>>
>>> On Sun, Nov 24, 2019 at 11:14:16AM +0100, Evilham wrote:
>>>> >- Evilham, please apply the patches to amprolla but with the
>>>> >ability to
>>>> > roll back if something goes pear-shaped.
>>>>
>>>>
>>>> Alright, will do during Hack4Glarus next weekend since I'll
>>>> be
>>>> online all
>>>> the time.
>>>
>>> Thanks for looking at this.
>>>
>>> We discussed amprolla last night at the meet, and I think that
>>> only 1 of the 2
>>> instances is doing anything useful. All http requests to
>>> packages.d.o are
>>> redirected to pkgmaster.d.o so the amprolla instance feeding
>>> packages.d.o could
>>> probably be disabled or removed. That only leaves 1 to update.
>>>
>>> Hope that helps.
>>
>> Thanks Mark, I had wondered but hadn't looked into that :-).
>>
>> Will apply the patch in the evening and keep an eye on things
>> over
>> the weekend, golinux briefly mentioned that there were other
>> amprolla-related concerns but they are fully independent of
>> this
>> particular fix.
>>
>> Just added a task to my queue to see if checksums can be an
>> issue
>> as mentioned on the notes, hopefully I also manage to do that
>> during the weekend :-).
>
> FTR: after reviewing the actually straightforward patch again
> with
> others, it has been deployed for quite a while now on the
> instance
> that affects pkgmaster.d.o (and therefore deb.d.o).
>
> Deployment was simple and unsurprising:
> 1. disable cronjobs
> 2. wait for cronjobs to finish
> 3. backup
> 4. apply patch
> 5. run orchestrate and full merge manually + test locking for
> the
> n-th time
> 6. re-enable cronjobs

Also FTR: several hours ago LeePen noticed a hard to spot typo
(more like a "apply programmer logic in shell" mistake) that
resulted in amprolla not syncing to pkgmaster; since no packages
had changed I hadn't noticed yet (but would have, and onefang
would have too).
The Merge Request is updated with that.

TL;DR: thanks to LeePen's discovery we are now even more confident
that the patch is alright and there is actual evidence of
synchronisation working, before there was absence of failure.

Cheers,
--
Evilham