Missing explicit check for workflow status failure

[ventsislav-georgiev]

The workflow status is documented to have 14 unique values: completed, action_required, cancelled, failure, neutral, skipped, stale, success, timed_out, in_progress, queued, requested, waiting, pending
Ref: https://docs.github.com/en/rest/actions/workflow-runs?apiVersion=2022-11-28

From looking at the code:

github_exporter/pkg/exporter/workflow.go

Lines 273 to 304 in d3cef39

	func conclusionToGauge(conclusion string) float64 {
	switch conclusion {
	case "completed":
	return 1.0
	case "action_required":
	return 2.0
	case "cancelled":
	return 3.0
	case "neutral":
	return 4.0
	case "skipped":
	return 5.0
	case "stale":
	return 6.0
	case "success":
	return 7.0
	case "timed_out":
	return 8.0
	case "in_progress":
	return 9.0
	case "queued":
	return 10.0
	case "requested":
	return 11.0
	case "waiting":
	return 12.0
	case "pending":
	return 13.0
	}
	return 0.0
	}

There is no check if the conclusion is failure and the only way to match against that is to use 0 as a failure status.
Is this expected?

[tboerger]

Yes, 0 should be used for failure. This is something that needs to be properly documented.

[ventsislav-georgiev]

Something seems to be wrong with the status reported. I'll have to debug for details but it seems that when the workflow is In Progress on Github, getting the metrics from github_exporter reports 0 for github_workflow_status. It then changes to 7 when the workflow completes.

[ventsislav-georgiev]

@tboerger I've added the following log:

And testing by running a workflow and querying the /metrics endpoint from a local build. Doesn't matter if the Github status is Queued or In Progress, the log is:

{"collector":"workflow","conclusion":"","level":"info","msg":"Returned 0 as conclusion","ts":"2023-07-12T16:39:24.072135Z"}

It seems the lib used for querying the API returns and empty string in this case...

[ventsislav-georgiev]

The label status (record.GetStatus()) seems to be the one with the expected queued, in_progress and other values. Conclusion returns value after the workflow is finished and it cannot be in most states that are cases in the switch.

[ventsislav-georgiev]

@tboerger I've pushed a small PR for the above scenario. Let me know what you think.

[tboerger]

Please give it a try with the latest tag of the image, this should work now. After getting it confirmed by you I would tag the new release v2.3.0

[ventsislav-georgiev]

@tboerger Thanks! Tried the latest tag and it works as expected, scraping Queued and In Progress states.

[ventsislav-georgiev]

@tboerger will you have time to push the v2.3.0 release?

[tboerger]

Since nobody complained about it I think I will roll a release after merging #236 and updating the changelog.

[tboerger]

I have just pushed v2.3.0, will be available in a few minutes.