Using the flannel CNI plugin for pod networking

The flannel CNI plugin provides pod networking without using IP addresses from a VCN's CIDR block. Communication between pods is encapsulated in the flannel overlay network, a simple private overlay virtual network that satisfies the requirements of the Kubernetes networking model by attaching IP addresses to containers. The pods in the private overlay network are only accessible from other pods in the same cluster. For more information about flannel, see the flannel documentation.

The flannel overlay network uses its own CIDR block to provision pods and worker nodes with IP addresses. The flannel CIDR block is not shared between clusters, so you can specify the same flannel CIDR block for multiple clusters. The default flannel CIDR block is large enough to support 65,534 pods and 512 worker nodes, and can be increased in size to support many more. Moreover, the number of pods per worker node is not determined by the node shape. Therefore, consider using the flannel CNI plugin if the density of pods per node presents an obstacle to the use of the OCI VCN-Native Pod Networking CNI.

You can use the flannel CNI plugin with managed node pools but not with virtual node pools.

In releases prior to July 2022, the clusters you created with Container Engine for Kubernetes always used the flannel CNI plugin for pod networking. In releases after July 2022, the OCI VCN-Native Pod Networking CNI plugin is the default if you use the Console to create clusters (running Kubernetes 1.22 or later). However, when using the 'Custom Create' workflow, you can choose to create clusters that use the flannel CNI plugin. The flannel CNI plugin continues to be the default if you use the API.