metrics for nodes / queue / builds

hashar commented 11 years ago

It would be nice to report a few more metrics to statsd. The jenkins-to-graphite python script at https://github.com/katzj/jenkins-to-graphite is meant to run via a cronjob, it poll the Jenkins API to report:

the queue size
builds started last minute / hour (could be done via a Statsd gauge)
executors (total, busy, free)
nodes (total, offline, online)

It can optionally log metrics on a per label basis if passed via --labels, would be nice to even support any label. The metrics for them are:

number of tied jobs
number of nodes having the label
number of executors (total, busy free)

Another option is reporting metrics on a per view based (same as labels, its optional and view names are to be passed via --jobs). For each view it would report:

of jobs in ok (blue), fail (red), warning (yellow) state
total # of jobs in the view

Not sure how easy it is to query those data with the Java internal API. But would definitely be nice to have :-]

hashar commented 11 years ago

Wikimedia downstream bug report used to track this issue: https://bugzilla.wikimedia.org/55988

joemiller commented 11 years ago

This would be nice to incorporate. I would gladly review and likely accept any patches that added additional metrics.

It was time consuming for me to find the right place to hook jenkins to get the current set of metrics. Getting these additional metrics may also be challenging and possibly require hooking into different points of jenkins. If any has sample code or suggestions on an approach, please share them here!

joemiller / jenkins-statsd-plugin

metrics for nodes / queue / builds #5

of jobs in ok (blue), fail (red), warning (yellow) state