Advice on recording rule "interval" when working around subquery range limitation

[franck102]

Hi all,

I am trying to record hourly increases over a metric obtained using a subquery to merge cAdvisor metrics with kube-pol-labels metric from kube-state-metric:

container_cpu_usage_seconds_total
* on(pod) group_left(<pod labels I want>) kube_pod_labels{<some filtering based on pod labels>}

Due to https://github.com/prometheus/prometheus/issues/1227 I can't just use 

increase(<subquery above>[1h])

 

and the advice is to use a recording rule to record the subquery.

I can't figure out how to configure the rule interval though? Conceptually what I would like is "generate one sample for every new sample received since the last rule evaluation"... is there any way to achieve this either explicitly, or implicitly?

Thanks!

Franck