Julien

Kubernetes implementation is the way to go, but should be behind a flag if possible.

You probably have resets. Can you please provide the raw metrics? to do that, run `mymetric{label="a"}[3h]` in the "console" tabs.

You have a counter reset: 439 @1617442554.304 439 @1617442554.315 440 @1617442564.304 439 @1617442564.315 440 @1617442574.33 440 @1617442584.297

It might be a bug in prometheus. Why is that scraped twice at the same time? Can you are more about your configuration?

I totally agree and I digged multiple times in the codebase without finding the bug yet...

Can you still provide the requested info please?

I'd like the prometheus.yml to see the relabeling rules and the jobs, if possible.

``` RichiH: Hysteresis: Allow a secondary PromQL query to disable alerts CONSENSUS: We intend to support hysteresis based on both query and/or cooldown time ```

I have open a pull request: #11827

hello, I think this is a breaking change in Prometheus, so not doable as-is in Prometheus 2.x. We would need to think about alternatives, like adding an extra "formatting" parameter...