[BUG] External metric values incorrectly interpreted by HPA

[ctpeter]

Agent version

7

Bug Report

What happened

We are using HPA with an external metric from Datadog via DatadogMetric CRD.

Example metric query:

max:kubernetes.cpu.usage.total{cluster_name:xxxxx, kube_deployment:xxxxx}.rollup(max,60)

HPA displays the metric as:

650634457m / 384M # CPU

Datadog shows the current value as 650634.4566410647

What we expected

• HPA should interpret the external metric value in the correct units (nanocores → millicores)
• The CURRENT column in kubectl get hpa should match the Datadog metric scale

Reproduction Steps

No response

Agent configuration

No response

Operating System

No response

Other environment details

No response

[vboulineau]

Hello,

When using external metrics, DatadogMetric returns the raw value (so if the metric is in nanocores, you get the nanocore value).

Either you need to put your HPA target in nanocores or add a / 1000000 in your Datadog query. The HPA has no notion of units, HPA will also always display values as milli values (as this is how Quantity object works, so 100.50 will always be 100500m