Resolve Metric issues, : poller don't get metrics correctly

[Neylix]

Describe the problem you discovered

• Metric should include all txns
• when timeout, metric dashboard should have metrics have of node itself.
  

Also on the dashboard it's seems that the datas are additionned instead of being averaged

Describe the solution you'd like

Metric poller should retrieve datas without error.
Dashboard view should have accurate values

[apoorv-2204]

Any way to reproduce it ? @Neylix

[internet-zero]

Hey team! Please add your planning poker estimate with ZenHub @apoorv-2204 @Neylix @prix-uniris

[ghost]

Also on the dashboard it's seems that the datas are additionned instead of being averaged

I think this is coming because the metrics exposed are not within the interval of 5s but global. The distribution/histogram metrics expose insights but without aggregate by scrape interval.
Instead it's the aggregating tool which should do it.
The problem is the _sum and _count metrics are increasing over time which is causing the increase line in the graph.
We should either:

• implement a kind of rate (as in prometheus: https://prometheus.io/docs/prometheus/latest/querying/functions/#rate)
• implement a kind of increase (https://prometheus.io/docs/prometheus/latest/querying/functions/#increase)

For example something like: increase[5s] = count at 10s - count at 5s

[internet-zero]

The issue estimate has been increased to 21, due to the increase in work

[samuelmanzanera]

Closed by #590