smartd needs to be restart on several occasions, including on failover.

Restarting smartd currently takes ~30 secs on a system with 400 disks, which is not acceptable.

It seems we issue smartctl commands for every disk, every single time. That doesnt scale.

We need to decrease the usages of smartctl as much as possible and only call them when necessary.

A suggested improvement would be to call it once when we boot or a disk is plugged in and cache whatever is needed.

After some investigation from Caleb the _service_changed method could also be rethinked, replacing it with service.started() made it a lot faster.

Another consideration is that for TrueNAS hardware we dont care about odd smartctl args, or putting disks to sleep, so these routines can be completely skipped for our hardware.

We have a M60 with 400+ disks to use for this for a short period of time. I will reach out via Slack for creds.