Thundering herd prevention? about sql_exporter HOT 1 CLOSED

Hey @reubano, given the design of this particular sql_exporter (it's synchronous, i.e. it's triggered by Prometheus requests), we want requested results as soon as possible. So I don't think we can introduce jitter. I did some research on that (branch fix/thundering-herd link) by adding some random sleep between goroutine executions. Even though it works, it requires aligning with Prometheus request timeouts, local sql_exporter timeouts, etc. In my opinion, it doesn't seem to be a solution for the time being.

Instead, I think we can solve this problem differently by configuring multiple jobs (with specific intervals and other useful settings) from Prometheus. So you can group your queries in sql_exporter jobs and call them based upon your own schedule via query string. The prototype is ready, and I plan to release the feature with the next v0.14 version.

Related issue is here: #178

from sql_exporter.