-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow cpu/memory scaler to target a specific container in the pod #1378
Comments
ref:kubernetes/kubernetes#90691 Maybe we can add this support in subsequent versions, when this feature is released in kubernetes 1.20 |
Oh - my impression was that Keda implements its own cpu/memory scaler and thus wouldn't be bound by the limitations in Kubernetes (which now seem to be mitigated). |
NO, cpu/memory scaler depend k8s HPA. keda is not implemented separately. |
Jup, we just provide a user-friendly scaler on top of it so that you can use the same experience for CPU/Memory as if it would be another scaler. |
makes total sense. Thanks for explaining! I guess we can close this. |
Getting error on EKS fargate cluster |
Please share the fix for this issue |
Use-Case
One limitation with Kubernetes' default hpa implementation, is that scaling happens based on the "sum" of all containers in the targeted pod. This means that sidecar metrics affect the scaling, and thus sidecar cpu/memory requests/limits have to be very carefully managed in order to achieve the desired scaling dynamics. Especially in cases where sidecar cpu/memory resources are tiny compared to those of the "main" app container, tuning this correctly is often very difficult. The kubernetes metrics-server actually presents per-container metrics, it's just that the default hpa doesnt have a way of selecting which individual container to track metrics for.
Since keda implements its own cpu/memory scaler, it would be so awesome if a new optional field
container
was added, which allowed using only the metrics of a specific container (ignoring other containers in the pod) for scaling.The text was updated successfully, but these errors were encountered: