Allow running Jobs to mark themselves as suspended

At the top level I think we want to take the job out of the queue
entirely and requeue it when the series is unfrozen.

On Thursday 26 May 2011 22:23:15 you wrote:
> At the top level I think we want to take the job out of the queue
> entirely and requeue it when the series is unfrozen.

That's not the right thing to be doing. The job is held until an archive
admin approves the package, at which point the job is released.

That's orthogonal to distroseries status, where even regular uploads are not
automatically released when its status changes.

On Fri, May 27, 2011 at 10:01 AM, Julian Edwards
<email address hidden> wrote:
> On Thursday 26 May 2011 22:23:15 you wrote:
>> At the top level I think we want to take the job out of the queue
>> entirely and requeue it when the series is unfrozen.
>
> That's not the right thing to be doing.  The job is held until an archive
> admin approves the package, at which point the job is released.
>
> That's orthogonal to distroseries status, where even regular uploads are not
> automatically released when its status changes.

Ah, I think there is a layering confusion here.

At the package layer we certainly need to know that there is something
waiting for an archive admin to approve.

At the queue layer we don't want stuff in a queue that can't be
processed: they shouldn't be queued for handling by the job runner.

-Rob

On Friday 27 May 2011 04:40:44 you wrote:
> Ah, I think there is a layering confusion here.
>
> At the package layer we certainly need to know that there is something
> waiting for an archive admin to approve.
>
> At the queue layer we don't want stuff in a queue that can't be
> processed: they shouldn't be queued for handling by the job runner.

Why not?

The PackageCopyJob holds all the context.

On Fri, May 27, 2011 at 8:10 PM, Julian Edwards
<email address hidden> wrote:
> On Friday 27 May 2011 04:40:44 you wrote:
>> Ah, I think there is a layering confusion here.
>>
>> At the package layer we certainly need to know that there is something
>> waiting for an archive admin to approve.
>>
>> At the queue layer we don't want stuff in a queue that can't be
>> processed: they shouldn't be queued for handling by the job runner.
>
> Why not?

Its incompatible design wise with zeromq/rabbitmq etc. Stuff in a
queue needs to be actionable.

-Rob

On Friday 27 May 2011 09:22:11 you wrote:
> Its incompatible design wise with zeromq/rabbitmq etc. Stuff in a
> queue needs to be actionable.

Why do we have a suspended status then?

Surely if something is suspended it's not going to be in a rabbit queue.

Fixed in stable r13128 <http://bazaar.launchpad.net/~launchpad-pqm/launchpad/stable/revision/13128>.