You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What problem this enhancement will solve?
External traffic needs to reach the deployed service in the edge device, hence NodePort service is needed. We need to ensure that the implementation of this function must consume optimal CPU/Mem resources, given that resources constrained edge device is the target for this CNI. The definition of "optimal" at this point of time is ensure that the cpu/mem requirement is lesser than bridge-cni and flannel CNI.
What is your proposal?
Please update the final proposal once it's ready for implementation.
Describe alternatives you've considered
KPNG (Kube Proxy Next Gen) eBPF implementation can be a possible candidate for this implementation. We need to ensure that integrating that implementation with Patu CNI doesn't spike the resource requirement of the overall CNI.
Additional context
At this point of time, we can take an assumption while implementing this feature, that all the endpoints of the backend cluster-ip services for the Ingress will be within the node, and not deployed across multiple nodes.
The text was updated successfully, but these errors were encountered:
What problem this enhancement will solve?
External traffic needs to reach the deployed service in the edge device, hence NodePort service is needed. We need to ensure that the implementation of this function must consume optimal CPU/Mem resources, given that resources constrained edge device is the target for this CNI. The definition of "optimal" at this point of time is ensure that the cpu/mem requirement is lesser than bridge-cni and flannel CNI.
What is your proposal?
Please update the final proposal once it's ready for implementation.
Describe alternatives you've considered
KPNG (Kube Proxy Next Gen) eBPF implementation can be a possible candidate for this implementation. We need to ensure that integrating that implementation with Patu CNI doesn't spike the resource requirement of the overall CNI.
Additional context
At this point of time, we can take an assumption while implementing this feature, that all the endpoints of the backend cluster-ip services for the Ingress will be within the node, and not deployed across multiple nodes.
The text was updated successfully, but these errors were encountered: