Unfortunately still a bit buggy, should be resolved in later Grafana (?)
Let's see how it looks with more than one host (sorry, on production only) ...
One dash for actual backup durations on the vertical (y) axis
Gets daily user ciphertext backup duration in seconds from Loki, transforms to number format, and graphs
Paint green from 0s, then yellow from 3600s (reason: that's a long time), then red from 10800s (reason: that's the first chance it could collide with the borgbackup-check-repo job)
Hopefully looks nice and gives quick info in absolute numbers
"duration" input sadly not available for borgbackup "check" job. That's why we also have:
Two dashes that graph the systemd unit's job states on the horizontal (x) axis:
Shows backup jobs being active over time
Nice to see if jobs overlap (not great if all hosts run backups at the same time)
Using Prometheus -> comes with alerting if backup job state is "active" for too long
