692804 - we should be able to tell slavealloc to send a specific number of slaves of a given pool to a specifc master

Status: RESOLVED WONTFIX

(Release Engineering :: General, defect, P4)

Description

[John Ford [:jhford] CET/CEST Berlin Time]

When doing a time-critical release, its really frustrating to see a bunch of jobs pending on 1 of 2 slaves.  We should have a knob that says "I am doing a release and I want X of this type of slave without having to specify specific slaves to push around

Comment 1

[Armen [:armenzg]]

IMHO the right way is to allow release jobs to happen in any master rather than lock them to a specific master where we really have a silo rather than the whole pool.

Comment 2

[[e-mail] (:bhearsum)]

I think this is a great idea. Let's be sure to avoid calling turning this into "release mode" when we build it, though. Having the generic ability to say "gimme more slaves" could be useful elsewhere.

Comment 3

[Chris AtLee [:catlee]]

(In reply to Armen Zambrano G. [:armenzg] - Release Engineer from comment #1)
> IMHO the right way is to allow release jobs to happen in any master rather
> than lock them to a specific master where we really have a silo rather than
> the whole pool.

actually, I'd rather have masters that *only* did releases.

Comment 4

[John Ford [:jhford] CET/CEST Berlin Time]

(In reply to Chris AtLee [:catlee] from comment #3)
> (In reply to Armen Zambrano G. [:armenzg] - Release Engineer from comment #1)
> > IMHO the right way is to allow release jobs to happen in any master rather
> > than lock them to a specific master where we really have a silo rather than
> > the whole pool.
> 
> actually, I'd rather have masters that *only* did releases.

Yes.  There are other advantages to having masters that only do releases, like being able to use a snapshot of releng infra that has been vetted by a staging release + manual testing.

Comment 5

[[e-mail] (:bhearsum)]

(In reply to Chris AtLee [:catlee] from comment #3)
> (In reply to Armen Zambrano G. [:armenzg] - Release Engineer from comment #1)
> > IMHO the right way is to allow release jobs to happen in any master rather
> > than lock them to a specific master where we really have a silo rather than
> > the whole pool.
> 
> actually, I'd rather have masters that *only* did releases.

Yeah, this is good to have, too. I'm having trouble coming up with a specific use case, but I still feel like this feature could be useful.

Comment 6

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

Found in triage.

Unclear from the back-forth in this bug as to what the actual objective is here, but regardless, I *think* this is the correct component.

Comment 7

[Chris Cooper [:coop] (he/him)]

WONTFIX in favor of bug 1135818.