Allow use of other metrics-adapter together with keda-metrics-adapter

[frank-berlin]

Allow use of other metrics-adapter like custom-metrics-stackdriver-adapter together with keda-metrics-adapter.

There is the current limitation to k8s that only one custom and external metrics-adapter can be used.
To workaround the problem, the keda-metrics-adapter could forword requests what there can't handle to an other metrics-adapter.
For the discovery calls it would need to do a merge with the keda data.
A clear and concise description of what you want to happen

Use-Case

Use keda and "non keda" HPA in same cluster based on external metrics.
For example custom-metrics-stackdriver-adapter and use metrics from appengine, memcache,...
Other usage is to allow a smooth migration of of existing HPA to keda based HPA, instead need todo all in one step.

This also reduce the scalers need to implemented, as sources available via other external metrics-adapter can be used for HPA.

Specification

filter request which are for keda metrics and process them as normal. All other need to forwarded other external metrics-adapter.
For discovery requests:

• forward to other external metrics-adapter
• do internal discovery
• merge results

[markusthoemmes]

How would keda‘s metric adapter know about the other metric adapters in the system though? Can‘t there only be one apiservice registered?

[frank-berlin]

This need to be configuration at keda-metrics-adapter.
Also the apiservice register of other metrics-adapter, need to removed/overwritten by keda-metrics-adapter.
From k8s view keda-metrics-adapter is visible only.

[zroubalik]

This could potentially result into never ending overwriting of the APIService if the other metrics-adapter is for example deployed by an operator, that watches it's resources. I am not sure if overwriting resources owned by another app is a good practise 🤷‍♂️

But interesting idea, @frank-berlin do you have any example of how could be done the forwarding of unknown metrics to the other metrics-server?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

[stale]

This issue has been automatically closed due to inactivity.