FR: Smooth out launching of smartctl

[eatnumber1]

On my machine I have hundreds of /dev/sd* devices. With smartctl_exporter, that means that when it wakes up it launches hundreds of smartctl commands at once, causing significant contention and noticeable hitches on the machine.

Could you add a feature that smooths out launching of smartctl commands? For example, if --smartctl.interval=60s, and the machine has 60 drives, it could launch 1 smartctl command every second.

[k0ste]

On my machine I have hundreds of /dev/sd* devices. With smartctl_exporter, that means that when it wakes up it launches hundreds of smartctl commands at once, causing significant contention and noticeable hitches on the machine.

Could you add a feature that smooths out launching of smartctl commands? For example, if --smartctl.interval=60s, and the machine has 60 drives, it could launch 1 smartctl command every second.

I'll tell you a secret: if you poll 60 devices sequentially at 60-second intervals (smartctl.interval=60s), you might not reach the end of the list within 60 seconds. What should you do then?

You can use release 0.13.0 - it works sequentially, because before merge #204

[SuperQ]

Hmm, that PR was supposed to be implemented with an error group. Right now there's no way for the user to limit concurrency.

With error group, you can set a limit.

[eatnumber1]

If it doesn't finish within 60s of the start time (an overrun) you can either skip the next one or start two. Cron and systemd both do the former (skip).