[Infra Monitoring UI] Use "fit" function in Stack Monitoring graphs to communicate gaps in data

[miltonhultgren]

If there is data missing due to connection issues or cluster overload, we get graphs with gaps like this:

We could make use of the fit function of the charts library to generate graphs like this instead:

[elasticmachine]

Pinging @elastic/infra-monitoring-ui (Team:Infra Monitoring UI)

[miltonhultgren]

Some more resources:
elastic/elastic-charts#1505
https://github.com/markov00/elastic-charts/blob/835bb69e00a2e905d6b11d766207fb5a97d2ab62/storybook/stories/mixed/6_fitting_stacked.story.tsx#L205-L224
https://elastic.github.io/elastic-charts/?path=/story/mixed-charts--fitting-functions-non-stacked-series&globals=theme:light

[markov00]

@miltonhultgren @formgeist we, the datavis team, are reconsidering our thoughts on gap handling in charts, so I think it is a good issue to discuss about that topic.
The topic says communicate gaps in data and I like to ask you few questions about that:

• how important are gaps with respect to the represented trend?
• what are the different causes that, in this specific use case, can cause a gap to appear? are all at the same level?
• I don't have enough knowledge about the topic, but looking at the chart we are talking about the indexing rate. If we have a gap, caused by a cluster overload, can we consider that indexing rate at zero? or it depends by other factors and, for example, the indexing rate is just unknown because the agent (or what is monitoring that rate) is unresponsive but we are still indexing the data?
• do we want to be really precise and show exactly where those gaps are or do we want just a global warning about that fact?
• ES has gaps policy, can be applied in your case?

[miltonhultgren]

I'll defer most questions to our UX experts but in most cases I would say that the gap happens because the agent is having issues (either with collection or delivery), so the indexing rate is definitely unknown (not 0). So the gap tends to highlight an issue with the agent, and not an issue with the monitored system, so a global warning within the chart might be more accurate.
I'm not aware of what ES gap policies are.

[markov00]

@miltonhultgren the current gap policies are described here https://www.elastic.co/guide/en/elasticsearch/reference/current/search-aggregations-pipeline.html#gap-policy
Basically you can mitigate those gaps by using zero, skipping the bucket or use a static value instead.
If the case is that the indexing rate is unknown then we should definitely understand if what is the importance of that information in relation to the overall monitoring trend, to avoid focusing on problems that are not primary for your user.

[miltonhultgren]

Since we mostly care about communicating unknown values in this use case, it feels like gap policies won't help since reporting 0 is wrong and skipping also feels wrong as it would hide the gap?

[pmeresanu85]

Added the enhancement label on this one