Scrape timeout with many disks

[pokab]

I manage a Linux box with around 20-25 HDDs. Some of these disks are faster to reply to smartctl, others are pretty slow, taking around 1-2 seconds. Setting the scrape interval to 60 seconds and scrape timeout to 40 seconds does not help in avoiding regular scrape timeouts. A previous solution (not specific to Prometheus) I made spawned the smartctl subprocesses in parallel for all the HDDs, and it works perfectly. Would a solution like this be appropriate for this software? Maybe with an option to enable or disable it?

[SuperQ]

This should be possible here with a goroutine worker pool. We can parallellize the data collection.

It doesn't look like the current collector records any timing information, even in debug mode. Something we can improve as well.

[pokab]

I've been using my forked version (see https://github.com/prometheus-community/smartctl_exporter/pull/204) for three weeks without any problem. It successfully solved the scrape issue. Could you please give some feedback?