Prometheus Metrics about bazel-buildfarm HOT 6 CLOSED

I would be really interested in gathering input on what metrics are useful in your experiences; we spawned a mailing list discussion on this today on the BuildGrid mailing list. https://lists.buildgrid.build/pipermail/buildgrid/2018-October/000035.html

CC: @t-chaik

from bazel-buildfarm.

Metrics are definitely in my sights - the logging output I've put together in #191 will cover some of the execution side, and is tractable by tools like fluentd with some basic scraping. I'm completely unfamiliar with prometheus and what would be beholden upon this project to provide useful metrics for it.

That is very high praise from you, sir. Thanks! We strive for stability mostly through inaction.

from bazel-buildfarm.

Taking into account a excerpt from the BuildGrid thread metioned above:

From the BazelCon discussions, the consensus seems to
be that the monitoring subsystem should be pluggable so that anybody
would be able to write a plugin for the monitoring system he cares about.

@werkt What do you think about using dropwizard metrics library which is backend agnostic so that anyone can plug its own metrics reporter (i.e. Prometheus, M3, Statsd, etc)?

CC @elianuber

from bazel-buildfarm.

I'm open to pull requests incorporating it. Hopefully that's the right level to report things so that there won't be substantial friction on integration.

from bazel-buildfarm.

opentelemetry-java could also work, maybe?

from bazel-buildfarm.

Added @80degreeswest to this, as he's indicated he's going to pursue metrics publishing via prometheus.

from bazel-buildfarm.