feature request: speedbump for publishing during freezes

[Seth Arnold]

Hello, as part of the release process, our release managers ask the security team and SRU members to not upload new packages into -updates or -security without coordination:

https://lists.ubuntu.com/archives/ubuntu-devel/2022-July/042227.html
https://lists.ubuntu.com/archives/ubuntu-devel/2022-February/041864.html

In this most recent freeze, this was overlooked several times. The security team added https://git.launchpad.net/ubuntu-qa-tools/commit/?id=4ec6109cdb45dd0582f8305d7f5f8fa2fd84836e to our tooling to try to discourage these mistakes from happening again, but this tooling may not be used by the kernel team.

So, I'd like to request a feature: a speed bump of some sort to discourage publishing to the -updates or -security pocket during these freeze events. While coordination would be the ideal, the habitual workflow of releasing updates is quite strong, and even when several of us knew about both the freeze and the nvidia coordinated release date, we didn't notice the conflict until after it had caused trouble for the release team.

Thanks

[Colin Watson (cjwatson)]

I'm concerned that adding a speed bump will just result in people getting used to doing the necessary thing to get past it, and then we'll be right back to where we were except with more entrenched complexity.

Leaving that aside, I'm not sure this should be done in Launchpad itself, in any case. Doesn't the kernel team use `sru-release` from lp:ubuntu-archive-tools? This feels like something best handled on the client side.

[Seth Arnold (seth-arnold)]

Ah that's a good concern. I would hope for a freeze of several days several times a year that it would be difficult to get used to bypassing it, but we are an amazingly adaptable species.

I'll try to get an answer on the kernel publication.

Thanks

[Andy Whitcroft (apw)]

Cirtainly any kind of block would need to be in the clients as well to stop us failing things, so it may as well be in the client. For us it would also need to be in our automation tooling to stop that asking. sru-release is the obvious home for that.

The main issue with a blanket stop (for us) would be that kernels are nominally independant, and the freezes typically are only required for seeded packages. This makes getting an appropriate block in place hard without over-blocking and leading to a natural "--unblock" always ending up on all your command lines.

In the case in of point-release breakage; we really shouldn't be releasing without coordination and for kernel we have a team member in the loop with the release-team throughout; typically they are waiting for regressions in kernels to be resolved before they can spin (sadly).

Will start some discussion on this regardless.

[Launchpad Janitor (janitor)]

[Expired for Launchpad itself because there has been no activity for 60 days.]