1388484 - should we reconsider our signoff policy in case of urgent No-updates?

Status: RESOLVED DUPLICATE of bug 1307169

(Release Engineering Graveyard :: Applications: Balrog (frontend), enhancement)

Description

[Mihai Tabara [:mtabara]⌚️GMT]

Today we had to turn off the updates for Firefox:release and it seems we're short on people for emergencies like that and it takes a bit of time to signoff, especially when this is super urgent.

Is there a way to granular this policy to reduce the signoff needed to just Releng + Relman? Is this something we want to do?

Comment 1

[Mihai Tabara [:mtabara]⌚️GMT]

Adding a NI on Ben.

Comment 2

[[e-mail] (:bhearsum)]

I agree this is a problem, and more urgent now that Multiple Signoffs is in place.

I think the best solution to this is https://bugzilla.mozilla.org/show_bug.cgi?id=1307169, which will allow a single person to shut off updates in a safe way (but not turn them back on).

Comment 3

[Sylvestre Ledru [:Sylvestre]]

Disabling updates won't have a big impact to the population. 
We should just make sure that the right persons are informed that it has been done.

Comment 4

[Chris AtLee [:catlee]]

(In reply to Sylvestre Ledru [:sylvestre] from comment #3)
> Disabling updates won't have a big impact to the population. 
> We should just make sure that the right persons are informed that it has
> been done.

What's an acceptable SLA for shutting off updates then? 1 hour? 4 hours?

Comment 5

[Mihai Tabara [:mtabara]⌚️GMT]

Dropping a NI to :Sylvestre for when's he's back from PTO - for comment above.

Comment 6

[Sylvestre Ledru [:Sylvestre]]

Maybe having a big red button for release mgmt to disable updates would be enough. This would have been helpful last Saturday.
Otherwise, an hour is fine I guess.

Comment 7

[[e-mail] (:bhearsum)]

We might be able to implement something hacky in the short term. Eg: bypass signoff if the only thing changing is mapping being set to No-Updates. This probably has some security holes, since someone could modify No-Updates when it's not mapped to, and then point at it without signoff. It might be worth the tradeoff depending on how painful this is.

Unless someone else has a better idea, I think the proper fix for this is bug 1307169.

Comment 8

[[e-mail] (:bhearsum)]

Does this bug have any value? I think https://bugzilla.mozilla.org/show_bug.cgi?id=1307169 is the clear fix for it, so we could just dupe it over.

Comment 9

[Mihai Tabara [:mtabara]⌚️GMT]

(In reply to Ben Hearsum (:bhearsum) from comment #8)
> Does this bug have any value? I think
> https://bugzilla.mozilla.org/show_bug.cgi?id=1307169 is the clear fix for
> it, so we could just dupe it over.

Makes sense. Let me do that, thanks!