Skip to content

Latest commit

 

History

History
95 lines (75 loc) · 4.26 KB

getting-started.md

File metadata and controls

95 lines (75 loc) · 4.26 KB

Getting started

Antrea is super easy to install. All the Antrea components are containerized and can be installed using the Kubernetes deployment manifest.

Ensuring requirements are satisfied

When using kubeadm to create the Kubernetes cluster, passing --pod-network-cidr=<CIDR Range for Pods> to kubeadm init will enable NodeIpamController. Clusters created with kubeadm will always have CNI plugins enabled. Refer to Creating a cluster with kubeadm for more information about setting up a Kubernetes cluster with kubeadm.

When the cluster is deployed by other means then:

  • To enable NodeIpamController, kube-controller-manager should be started with the following flags:

    • --cluster-cidr=<CIDR Range for Pods>
    • --allocate-node-cidrs=true

  • To enable CNI network plugins, kubelet should be started with the --network-plugin=cni flag.

As for OVS, when using the built-in kernel module, kernel version >= 4.4 is required. On the other hand, when building it from OVS sources, OVS version >= 2.6.0 is required.

Red Hat Enterprise Linux and CentOS 7.x use kernel 3.10, but as changes to OVS kernel modules are regularly backported to these kernel versions, they should work with Antrea, starting with version 7.4.

In case a node does not have a supported OVS module installed, you can install it following the instructions at: Installing Open vSwitch.

Installation

To deploy a released version of Antrea, pick a deployment manifest from the list of releases. For any given release <TAG> (e.g. v0.1.0), you can deploy Antrea as follows:

kubectl apply -f https://github.com/vmware-tanzu/antrea/releases/download/<TAG>/antrea.yml

To deploy the latest version of Antrea (built from the master branch), use the checked-in deployment yaml:

kubectl apply -f https://raw.githubusercontent.com/vmware-tanzu/antrea/master/build/yamls/antrea.yml

Deploying Antrea on a Cluster with Existing CNI

The instructions above only apply when deploying Antrea in a new cluster. If you need to migrate your existing cluster from another CNI plugin to Antrea, you will need to do the following:

  • Delete previous CNI, including all resources (K8s objects, iptables rules, interfaces, ...) created by that CNI.
  • Deploy Antrea.
  • Restart all Pods in the CNI network in order for Antrea to set-up networking for them. This does not apply to Pods which use the Node's network namespace (i.e. Pods configured with hostNetwork: true). You may use kubectl drain to drain each Node or reboot all your Nodes.

While this is in-progress, networking will be disrupted in your cluster. After deleting the previous CNI, existing Pods may not be reachable anymore.

For example, when migrating from Flannel to Antrea, you will need to do the following:

  1. Delete Flannel with kubectl delete -f <path to your Flannel YAML manifest>.
  2. Delete Flannel bridge and tunnel interface with ip link delete flannel.1 && ip link delete flannel cni0 on each Node.
  3. Deploy Antrea.
  4. Drain and uncordon Nodes one-by-one. For each Node, run kubectl drain --ignore-daemonsets <node name> && kubectl uncordon <node name>. The --ignore-daemonsets flag will ignore DaemonSet-managed Pods, including the Antrea Agent Pods. If you have any other DaemonSet-managed Pods (besides the Antrea ones and system ones such as kube-proxy), they will be ignored and will not be drained from the Node. Refer to the Kubernetes documentation for more information. Alternatively, you can also restart all the Pods yourself, or simply reboot your Nodes.

To build the image locally, you can follow the instructions in the Contributor Guide.

Antrea components can also be run manually as processes for development purposes. See Manual Installation for information.

Deploying Antrea in Kind

To deploy Antrea in a Kind cluster, please refer to this guide.