Hello Guix!
Over the last few days, out of frustration ;-), I hacked Cuirass to
improve several things:
• Logging is improved: useful events are logged, including build
started/succeeded/failed (using a variant of what I proposed in the
Guix ‘wip-ui’ branch). This makes it much easier to understand
what’s going on!
• Concurrency: the evaluate/build cycle would previously happen
sequentially. This was bad for latency (it could take a while
before we would pull from the repo) and for resource usage (while
evaluating all the build machines would be idle). Now there’s some
concurrency, with a switch to Fibers (yay!), which means that these
activities can be interleaved, as well as HTTP request processing.
• Build status: the starttime/stoptime of a ‘Build’ entry in the
database is now accurate since we set them once the build has
actually started/stopped. This means the database is updated as
soon as a build finishes, rather than when the whole
‘build-derivations’ RPC has returned. We can also distinguish
between started and scheduled builds now.
• HTTP API: as a corollary, /api/latestbuilds now really returns the
latest builds. There’s a new /api/queue that returns pending
builds—builds that have a database entry with ‘status’ < 0.
• Restarting unfinished builds: it’s common, especially when testing,
to interrupt Cuirass, leaving a number of builds unfinished or not
even started. Now Cuirass restarts those upon startup.
And! This brings a whole set of new bugs that I’m hunting notably on
berlin (which may thus lag behind…). Overall I think it’ll make Cuirass
easier to work with and more “introspectable”.
Feedback welcome!
Ludo’.
