Kubernetes custom-metrics-apiserver interface implentation, using Sydig Monitor as metrics backend.
-
Currently, the metric value is just set to
0if the backend doesn't respond. Other options:- Cache metrics and return last known value
- Fallback to CPU usage
- Return an explanatory error message and terminate the pod
-
Backend only implements some interface functions (the ones required by the HPA).
- Create a complete metrics provider implementation, including error handling and defensive code