jschwartzy

> I've also had a meeting with AWS CNI guys, and they suggested using CIDR prefix reservation to seperate /32 and /28 allocation to limit fragmentation. I'm not sure how...

After encountering this issue on EKS 1.28 and CA 1.28.2, we reproduced the issue as described above by manually removing one of the Kubernetes nodes. Taking a look at the...

We're also encountering a similar issue with Envoy via the `GatewayAPI`. We can narrow it down to when Envoy is on the name node as the destination pod, we hit...

Some additional observations, providing Hubble traces: 1. Envoy proxy is on a different node than the app replica. Note that all traffic is directed to the pod replica ``` Feb...

> I don't mind this in genera,l but a lot of those example annotations ought to be replaced by first class API fields that already exist Appreciate the feedback. Typically,...

Really appreciate the follow-up and suggestion. I attempted to comment out this code to observe the behavior: ``` // if len(option.Config.GetDevices()) == 1 { // if e.IsHost() || !option.Config.EnforceLXCFibLookup() {...

Issue is still present in `1.16.0`.

Attempted to include the PR linked above into Cilium `1.16.1` but no change to the behavior. If we comment out the code linked below, the co-located Envoy proxy/replica routing works...

Tested on `1.16.4` successfully, thank you for following up!