Is it possible do downgrade k3s?

[loxy]

I saw already discussion #376 and postet a comment there. But my question goes a little bit beyond, so I started a new thread.

I have a related issue (with same error log) because I have problems with postgres operator (PGO). It is not compatible with 1.25...

Is it possible to downgrade to version 1.24 without any problems?

So, that means that once v1.25.3+k3s1 was out, my cluster automatically switched to that Kubernetes version (while specifying stable)?

And if I set initial_k3s_channel = "v1.24" and apply again it will be reverted? With no drawbacks?

[otavio]

Considering we'd like to upgrade to v1.24 releases, we can do this using following file:

# kubectl apply -f ./k3s-upgrade.yaml
# Doc: https://rancher.com/docs/k3s/latest/en/upgrades/automated/
# agent plan
apiVersion: upgrade.cattle.io/v1
kind: Plan
metadata:
  name: k3s-agent
  namespace: system-upgrade
  labels:
    k3s_upgrade: agent
spec:
  concurrency: 1
  channel: https://update.k3s.io/v1-release/channels/v1.24
  serviceAccountName: system-upgrade
  nodeSelector:
    matchExpressions:
      - {key: k3s_upgrade, operator: Exists}
      - {key: k3s_upgrade, operator: NotIn, values: ["disabled", "false"]}
      - {key: node-role.kubernetes.io/control-plane, operator: NotIn, values: ["true"]}
  tolerations:
    - {key: server-usage, effect: NoSchedule, operator: Equal, value: storage}
  prepare:
    image: rancher/k3s-upgrade
    args: ["prepare", "k3s-server"]
  drain:
    force: true
    skipWaitForDeleteTimeout: 60
  upgrade:
    image: rancher/k3s-upgrade
---
# server plan
apiVersion: upgrade.cattle.io/v1
kind: Plan
metadata:
  name: k3s-server
  namespace: system-upgrade
  labels:
    k3s_upgrade: server
spec:
  concurrency: 1
  channel: https://update.k3s.io/v1-release/channels/v1.24
  serviceAccountName: system-upgrade
  nodeSelector:
    matchExpressions:
      - {key: k3s_upgrade, operator: Exists}
      - {key: k3s_upgrade, operator: NotIn, values: ["disabled", "false"]}
      - {key: node-role.kubernetes.io/control-plane, operator: In, values: ["true"]}
  tolerations:
    - {key: node-role.kubernetes.io/control-plane, effect: NoSchedule, operator: Exists}
    - {key: CriticalAddonsOnly, effect: NoExecute, operator: Exists}
  cordon: true
  upgrade:
    image: rancher/k3s-upgrade

[mysticaltech]

Ahh, found it, you have two yaml to pull down:

• For cp (start by this): kubectl get plan k3s-agent -n system-upgrade -o yaml
• For agent: kubectl get plan k3s-server -n system-upgrade -o yaml

Then you can combine it into one file by separating it with a ---, and kubectl apply -f <name-of-combined-file> -n system-upgrade to that file.

[mysticaltech]

But again, before, try with terraform alone with initial_k3s_channel="v1.24" in your kube.tf and apply that again, normally it should be enough!

[loxy]

I can confirm it worked, but only with kubectl method. When using terraform apply after changing channel terraform says:

No changes. Your infrastructure matches the configuration.

Terraform has compared your real infrastructure against your configuration and found no differences, so no changes are needed.

Thank you again. Everything is working again. Awsome!

[CC007]

Unfortunately you can't use this anymore to downgrade k3s since the 2023-07 releases: https://docs.k3s.io/upgrades/automated#downgrade-prevention

[mysticaltech]

@CC007 Thanks for sharing!