[Nix-dev] Hydra queue is in a weird state

[Vladimír Čunát]

Hi,
around 24h ago I merged some older staging commit, and Hydra somehow
hasn't recovered since.

Most of jobs should've been cached from staging - mainly except for
Haskell stuff because that got updated on master in the meantime - but
apparently those jobs are stuck in the queue, not being processed in any
way. The machines were reportedly idling a lot yesterday after that
(now they picked some subsequent jobs on staging).

--Vladimir

[Eelco Dolstra]

Hi,

Post by Vladimír Čunát
around 24h ago I merged some older staging commit, and Hydra somehow
hasn't recovered since.
Most of jobs should've been cached from staging - mainly except for
Haskell stuff because that got updated on master in the meantime - but
apparently those jobs are stuck in the queue, not being processed in any
way. The machines were reportedly idling a lot yesterday after that
(now they picked some subsequent jobs on staging).

Can you be more specific (e.g. build IDs)? I don't see anything wrong at the moment.

[Vladimír Čunát]

Post by Eelco Dolstra
Can you be more specific (e.g. build IDs)? I don't see anything wrong at the moment.

Hmm, me neither, not anymore. The number of queued jobs on trunk* has
also dropped to the number I would expected right after the merge and I
see no particular problem anymore. It's also possible I just got
confused by something.

Just to be sure, the binaries from staging are cached at least for a few
days, right? Even if they don't belong to the latest evaluation anymore?

--Vladimir

[Eelco Dolstra]

Hi,

Post by Vladimír Čunát
Just to be sure, the binaries from staging are cached at least for a few
days, right?

Right now, there is no garbage collection of the cache.nixos.org S3 bucket *at
all*, so binaries are kept indefinitely.