Minor changes

[arodrime]

• period: now -3h (default)
• /!\ Important CPU chart fix, as multiple cores/CPU report separately --> we need to use avg, not min of idle for calculations

[rustyrazorblade]

I'm not wild about having dashboards be 3 hours by default, I think 15 minutes is a lot more useful, more often. For those CPU graphs, let's show min, max and average. Trying to distill it to a single graph doesn't provide enough context to derive any meaning from it.

[arodrime]

I'm not wild about having dashboards be 3 hours by default, I think 15 minutes is a lot more useful, more often.

I was wondering about this, but I guess you are right, we both are.

• 15 minutes is nice when running in the context of tlp-cluster (and we can keep it here, it makes sense).
• For a production cluster I would say 3 hours is a nice balance between just seeing the last minutes (Spikes can seem to be huge, no history, ...) and seeing so much time in the past that you don't see what's happening in the last 5 or 15 minutes.

I'd say both are valid I think, for TLP cluster 15 minutes (or maybe slightly more? - like 30' / 1h) might make more sense.

For those CPU graphs, let's show min, max and average. Trying to distill it to a single graph doesn't provide enough context to derive any meaning from it.

Yes, as discussed internally, this makes sense to me too. I'll do that this week. I know it's not that big, but I've been quite busy lately :(.