Have more relaxed read/write alerts for physical hosts as they are
probably hypervisors and regular high read/writes are more common.
Also differentiate between physical and virtual hosts for IO alerts and
allow for hard disks to spend more time in IO.
The scrape seems to take around a second to complete and with the
configured timeout of 1s that failed half the time. Therefore use the
default, more relaxed scrape interval and timeout and have it be
reliable.
For now introduce node-exporter/hosts alert rules, which got taken from
https://samber.github.io/awesome-prometheus-alerts/rules
However with the labels removed from the description, since they don't
render correctly (at least in Telegram) and don't seem to provide much
value, as we render the labels in the notification anyway.
Also only have Telegram as the notification channel for now, as it was
the easiest to set up.
christian
June
Stefan Bethke