Skip to content

Latest commit

 

History

History
27 lines (21 loc) · 1.9 KB

migrate_to_calicovpp.md

File metadata and controls

27 lines (21 loc) · 1.9 KB

Calico-VPP nodes are entirely compatible with regular Calico nodes, meaning that there can be both VPP enabled and regular nodes in the same cluster. This allows to migrate existing clusters to the VPP dataplane with minimal disruption.

Prerequisites

Before attempting a migration to the VPP dataplane, ensure that your cluster is running the same version of Calico than the version we use in the latest release of Calico-VPP. The calico version is mentioned in the Calico-VPP release tag.

Migrating nodes to the VPP dataplane

Before you get started with the migration, you should define a Calico-VPP configuration for your cluster. The simplest way to do so in most cases is to use a Baremetal configuration.

The process to deploy VPP on some nodes in a Kubernetes / Calico cluster is the following:

  • Add an annotation to all the nodes in the cluster, such as calico-vpp: disabled
  • Edit the calico-node DaemonSet to run only on the nodes that have this annotation: 
    spec:
  nodeSelector:
    beta.kubernetes.io/os: linux
    calico-vpp: disabled
  • Create the VPP version of the node DaemonSet, and configure its node selector to select only the nodes with a different annotation, such as calico-vpp: enabled
  • Then start migrating the nodes. For each node that you want to deploy VPP on:
    • Remove all the running pods on the node with kubectl drain --ignore-daemonsets <node>
    • Change the node annotation from calico-vpp: disabled to calico-vpp: enabled
    • Wait for the calico-vpp-node pod to start
    • Resume the scheduling of pods to this node with kubectl uncordon <node>
    • The new pods that are scheduled on this node will benefit from VPP networking. After migrating the first node, validate that the pods scheduled on this node are functional before migrating additional nodes.