Prior to Rancher v2.5.6, the
agent did not have native Windows manifests on downstream clusters with Windows nodes. This would result in a failing
agent pod for the cluster.
If you are upgrading from an older version of Rancher to v2.5.6+, you can deploy a working
agent with the following workflow in the downstream cluster:
- Cordon all Windows nodes.
- Apply the below toleration to the
- Uncordon all Windows nodes.
- Delete all
agentpods. New pods should be created with the new toleration.
- Once the
agentpods are running, and auto-update is enabled for Fleet, they should be updated to a Windows-compatible
tolerations: - effect: NoSchedule key: cattle.io/os operator: Equal value: linux