Hi Netdata team,
Are there any plans to enable us to dynamically populate custom Netdata cloud dashboards?
For example, I have a custom dashboard which tracks system temperatures and I add a charts for each node (for example, if using a gauge representation). However, this is not nice to manage on a small scale and doesn’t scale at all on a larger number of machines where adding or removing new nodes has to be done manually.
What I imagine is creating a custom dashboard and being able to specify a chart type and a disjunctive (‘or’) list of metrics* along with a node-membership rule (e.g. by space or name).**
Can this be done already but perhaps I’ve missed it?
If not then I can raise a feature request on Github to hash things out.
*disjunctive because, as in the helpful example of temperature sensors: a machine may have many sensors and additionally different types of machines will have different sensors. I know which (different) sensor(s) I care about on my Dell vs HP vs Raspberry Pi machines.
A related issue concerns the fact that one might consider the default chart which plots all nodes to be more efficient than a gauge-chart per node. Even in that case the challenge exists because such charts can’t, in my opinion, handle those sorts of metrics well for the following reasons:
Firstly, there is the sheer number of metrics. Suppose that I have around 5 sensors per node; one can see that I don’t need many nodes before that chart becomes visually cluttered.
The second issue is that most of those charts default to an averaging of the metrics; on a single node that doesn’t make sense and across nodes it’s less helpful. We can now, thanks to your work, set custom settings for a chart so that, as in this case, I use the MAX value rather than the AVG. That’s what I do now but that then loses me the ability to see the difference between the highest value metric/sensor and the others.
I do care about the hottest sensor on a node but it doesn’t follow that it’s the only one I care about.
This is because the maximum temperature for that sensor S1 (e.g. say 80C) is not necessarily a healthy value for another sensor S2- specifically, if S2 is lower than S1 it may be that S2’s value is still too high even though S1’s value is acceptable.
Consider another scenario that I have:
http_check monitoring of websites (or ping). I use multiple Netdata agents in different locations to monitor multiple URLs on multiple servers. A default consolidated chart is not helpful of reasons already gone into. However, what would be extremely helpful is if I could build a dynamic chart that said something like:
where collect_module is http_check AND URL = acme.com
**In v2 I imagine being able to place a kind of meta-placeholder/shape/zone (imagine the shapes that appear when you drag and select using a rectangular image-cropping tool) into which dynamic charts would appear. That would allow a custom dashboard where there are multiple zones each with different population rules.
BTW, Happy to get feedback if my chart-fu needs work 