Seeing these kind of responses from Jon and other people in the Nix community, it seems like almost no one has these privileges. It is quite surprising that the release managers don’t automatically get these privileges, since being able to get stuff on Hydra passing seems to be a big chunk of the responsibility of the release managers.
I’ve wanted the ability to restart stuff on Hydra for helping with the Haskell subsystem in nixpkgs, but I’ve never been sure who to ask to grant me access.
IRC is probably the best way to get in contact with these individuals.
Not sure how “liberal” you want to be with something as critical as hydra…
However, I think it makes sense for release managers to be able to do hydra configuration since part of the release process is creating the release channel, changing the “stable” attr, and a few other hydra activities.
You wouldn’t happen to know if there is any sort of policy about how to give out access to Hydra? For instance, is it only people on the infrastructure team that have access to restarting builds on Hydra? Or is there a set of people that aren’t on the infrastructure team, but do have access to Hydra?
Maybe @zimbatm would know, since he’s on the infrastructure team (and active on Discourse).
It’s like more things on the infrastructure side, access has been granted based on some random events. We don’t have a uniform or documented policy I’m afraid.
In my view, release managers should get that privilege at minimum. And potentially a larger set of trusted members once they have been sworn in. I don’t fully understand the impact of restarting jobs myself and that should be clarified to me and any people that get job-restarting privileges.
In the meantime, let’s use this thread. Please @mention me with the stuck jobset and I will restart it in the next 24h.
Sometimes @vcunat or me stop certain jobs because of resources. Unfortunately we have no place to communicate why something has stopped. At times we mention it in the current Staging next PR, but typically not.
In another thread, Jon was saying this appears to be a configuration error, so it is possible restarting it won’t do anything without fixing the underlying configuration error:
Normally I wouldn’t post about this, but given that libreoffice is a relatively popular package, I think it might look bad if NixOS gets a bunch of new users from the 20.09 release, but then they have to compile libreoffice and can’t get it from the cache.