I would like aggregate a metric with the maximum number of another metric.
Metric 1:
| 0:00 | 0:01 | 0:02 | |
|---|---|---|---|
| Instance 1 | 1 | 1 | 1 |
| Instance 2 | 2 | 2 |
2 |
| Instance 3 | 3 |
3 | 3 |
Metric 2:
| 0:00 | 0:01 | 0:02 | |
|---|---|---|---|
| Instance 1 | 1 | 12 | 23 |
| Instance 2 | 1 | 14 |
14 |
| Instance 3 | 3 |
3 | 3 |
For example, the aggregated result from Metric 1 is (3,2,1).
The result using
max(metric2)
is (3,14,23) and it can be mapping to (instance3, instance2, instance1). I would like to apply this result's instance back to metric1.
Query
metric1 and on(instance) topk(1, metric2)
will return data as you want it.
Here:
topk(1, metric2) takes top 1 value of metric2 (for each timestamp independently). It is similar to max, but preserves labels of original metric.and on(instance) filters out results of metric1 based on presence of the respective label instance in the right hand side.