Analyze a period of usage on the metric server with krr

[hafx]

Hello,
I'm using a centralized Victoria-Metric to export all our metrics (from various setup).
I target this metric server with krr from my PC. However, these setups differ in terms of service versions in pods and the number of nodes per cluster.

Different scenario :

1. single node : ver 1.x
2. multi node : ver 1.x of the setup
3. multi node : ver 2.x of the setup
4. ...

We're seeking recommendations for each scenario without deploying additional Victoria-Metrics instances per scenario.

The problem
With the current version of krr we can't specify a period, we can only use --history_duration.

I mean :
We can't for example select : "from 10/20/23 10:00 AM to 10/28/23"
Consequently, we're compelled to gather data from 10/20/2023 to today. However, this timeframe (from 10/28 to today) might be a period when the setup isn't in use, potentially distorting our insights.

Describe the solution you'd like
Being able to select a specific usage period would be beneficial in this context.
For example :

• from 10/11/23 10:00 AM to 10/19/23

Are you interested in contributing a PR for this?
As for contributing to a PR for this feature, I'm uncertain if I'll have the time.

Is there a solution for this case ?
Thanks

[Natan Yellin]

Hi @hafx,
It's not something we support today. Does it work if you just use --history_duration?

if the workload (e.g. Deployment) names are unique, I suspect it'll work with --history_duration as we ignore time periods for which we have no data in the calculation.

Regarding the the other setup issues, would you be interested in jumping on a quick call and discussing your use case more?