# Kubernetes networking with Edera

## Kubernetes networking with Edera

For Edera-managed pods, Edera should seamlessly use your existing cluster CNI configuration to set up networking for Edera pods.

The following CNIs have been tested with Edera:

### Cilium

Both IPv4 and IPv6 mode supported.

⚠️

When using Cilium’s [kube-proxy-replacement](https://docs.cilium.io/en/latest/network/kubernetes/kubeproxy-free/#kubeproxy-free) with Edera, the in-pod socket-level loadbalancer must be disabled for pods by configuring Cilium with `socketLB.hostNamespaceOnly=true`.

See [upstream Cilium docs](https://docs.cilium.io/en/latest/network/kubernetes/kubeproxy-free/#socketlb-host-netns-only) for more details.

### Flannel

Both IPv4 and IPv6 mode supported.

### AWS VPC-CNI

Both IPv4 and IPv6 mode supported.

### ipvlan

L2, L3, and L3s modes are supported, under both IPv4 and IPv6.

Currently, Edera does not support Multus or CNI configurations that configure multiple links for a single pod.

If there is a CNI misconfiguration, the `protect-cri` systemd service should report the error and fail to start.

You can check the status of `protect-cri` with `sudo systemctl status protect-cri` on an Edera-enabled node.

Note that if your CNI setup uses nonstandard CNI plugin binary or configuration paths, you will need to update [/var/lib/edera/protect/cri.toml](https://docs.edera.dev/reference/cri-toml/_index.md) with the nonstandard paths, and restart the `protect-cri` service via `sudo systemctl restart protect-cri`.