maas-devel team mailing list archive

[Gavin Panella <gavin.panella@xxxxxxxxxxxxx>]

On 5 August 2014 11:20, Graham Binns <graham.binns@xxxxxxxxxxxxx> wrote:
> I've just landed a fix for bug 1351451 (Impossible to release a BROKEN
> node via the API). This is a problem that we (Raphers and I) saw on the
> Orange Boxes (probably as a result of Raphinator and Gavinator-induced
> hackery, but still).
>
> A question arose in Jools's review:
>
>> We probably need to talk about this with rvb.  I was under the
>> impression that marking in-use nodes as broken should keep their owner
>> as this is a way to know that it was being used and marked broken by a
>> user, not an admin.  Do we need to keep that distinction?
>
> Since mark_broken() already releases the node, ISTM that there's been a
> misunderstanding somewhere. Of course, the node activity log could have
> "Marked broken by Foo" in it to guide admins…
>
> Anyway, I've landed the branch as-is, but it'd be good for us to all get
> on the same page re: behaviour when marking broken.

There are different classes of "broken" as we modelled in the node
lifecycle last week. One of them is the Needs Attention "state", which
is less a status and more a flag, because the node can continue
operations. The others were proper states, iirc. We probably want to
change the mark_broken() and mark_fixed() calls to transition a node
into origin-state-specific broken states, if that makes sense. Or we
could model it all with flags. I don't know, but it's clear that
mark_broken() and mark_fixed() were forged in a simpler age.