Question about recentFiler=1d for getting executions

[davidcpell]

Hi @phsmith, I'm integrating rundeck_exporter (thanks for this!) and while looking at the code came across the use of 1d as a recentFilter when getting executions:

rundeck_exporter/rundeck_exporter.py

Lines 232 to 233 in 82b203a

	endpoint = f'/project/{project_name}/executions?recentFilter=1d&max={project_executions_limit}'
	endpoint_running_executions = f'/project/{project_name}/executions/running?recentFilter=1d&max={project_executions_limit}'

Can you explain the reasoning behind this value? I just want to make sure things will work as expected when Prometheus is scraping every minute. Will I be duplicating metrics?

[phsmith]

Hey @davidcpell!
Yeah, let me explain.

That point in the code retrieves the 20 (max => project_executions_limit = > 20 by default) most recent project executions during the day, or in the last 24h.

For example, the rundeck_project_execution_status metric shows 5 metrics every time the exporter is scraped, each one representing the status of the rundeck jobs, and value is set to 1 for the current status and 0 for the others. Every minute you can see these metrics, so yeah, you can get the same metrics values over a period of time, but it's purely for having more accurate results.

Don't know if I was clear enough, lemme know if not.

[davidcpell]

I'm still struggling to see how this works out in a case where we are scraping every minute but the exporter is looking back 24h for executions. For example, if there were 10 executions of a certain job in the last 24h and Prometheus is scraping every minute, would we not be adding 10 to the counter every time a scrape takes place even if no more executions are happening (until 24h have gone by and the API query no longer finds those 10 executions)?

Or a gauge example: let's say we had 100 executions evenly spread out over the course of 24h, and the next day there are 0 executions the whole day. At noon the next day (halfway through the day) wouldn't we still be sending a value of 50 to Prometheus? And it would look like the value went from 100 to 50 when actually there are 0 executions for those times on the 2nd day?

Sorry if I'm misunderstanding something either in the exporter or Prom itself!

[phsmith]

@davidcpell You're absolutely right!
I've just made a few changes. I added the parameter --rundeck.projects.executions.filter, default to 5n (5 minutes), so you can now change the behavior of the query.
What do you think?