Upgrading your installation

To upgrade your Knative components and plugins, run the kubectl apply command to install the subsequent release. We support upgrading by a single minor version number. For example, if you have v0.14.0 installed, you must upgrade to v0.15.0 before attempting to upgrade to v0.16.0. To verify the version number you currently have installed, see Checking your installation version.

If you installed Knative using the either the eventing-operator or serving-operator plug-ins, the upgrade process will differ. See the serving-operator upgrade guide and the eventing-operator upgrade guide to learn how to upgrade an install managed by the operators.

Before you begin

Before upgrading, there are a few steps you must take to ensure a successful upgrade process.

Identify breaking changes

You should be aware of any breaking changes between your current and desired versions of Knative. Breaking changes between Knative versions are documented in the Knative release notes. Before upgrading, review the release notes for the target version to learn about any changes you might need to make to your Knative applications:

Release notes are published with each version on the “Releases” page of their respective repositories in GitHub.

View current pod status

Before upgrading, view the status of the pods for the namespaces you plan on upgrading. This allows you to compare the before and after state of your namespace. For example, if you are upgrading Knative Serving, Eventing, and the monitoring plug-in, enter the following commands to see the current state of each namespace:

  1. kubectl get pods --namespace knative-serving
  1. kubectl get pods --namespace knative-eventing
  1. kubectl get pods --namespace knative-monitoring

Upgrading plug-ins

If you have a plug-in installed, make sure to upgrade it at the same time as you upgrade your Knative components. For example, if you have the monitoring plug-in installed, upgrade it alongside Knative Serving and Eventing.

Run pre-install tools before upgrade

In some upgrades there are some steps that must happen before the actual upgrade, and these are identified in the release notes. For example, upgrading from v0.15.0 to v0.16.0 for Eventing you have to run:

  1. kubectl apply --filename https://github.com/knative/eventing/releases/download/v0.16.0/pre-install-to-v0.16.0.yaml

Upgrade existing resources to the latest stored version

Our custom resources are stored within Kubernetes at a particular version. As we introduce newer and remove older versions you’ll need to migrate our resources to the designated stored version. This ensures removing older versions will succeed when upgrading.

For the various subprojects - we have a K8s job to help operators perform this migration. The release notes for each release will explicitly whether a migration is required.

ie.

  1. kubectl create --filename https://github.com/knative/serving/releases/download/v0.16.0/serving-post-install-jobs.yaml

Performing the upgrade

To upgrade, apply the .yaml files for the subsequent minor versions of all your installed Knative components and features, remembering to only upgrade by one minor version at a time. For a cluster running v0.15.2 of the Knative Serving and Eventing components and the monitoring plug-in, the following command upgrades the installation to v0.16.0:

  1. kubectl apply --filename https://github.com/knative/serving/releases/download/v0.16.0/serving.yaml \
  2. --filename https://github.com/knative/eventing/releases/download/v0.16.0/eventing.yaml \
  3. --filename https://github.com/knative/serving/releases/download/v0.16.0/monitoring.yaml

Run post-install tools after the upgrade

In some upgrades there are some steps that must happen after the actual upgrade, and these are identified in the release notes. For example, after upgrading from v0.15.0 to v0.16.0 for Eventing you should run:

  1. kubectl apply --filename https://github.com/knative/eventing/releases/download/v0.16.0/post-install-to-v0.16.0.yaml

Verifying the upgrade

To confirm that your components and plugins have successfully upgraded, view the status of their pods in the relevant namespaces. All pods will restart during the upgrade and their age will reset. If you upgraded Knative Serving, Eventing, and the monitoring plug-in, enter the following commands to get information about the pods for each namespace:

  1. kubectl get pods --namespace knative-serving
  1. kubectl get pods --namespace knative-eventing
  1. kubectl get pods --namespace knative-monitoring

These commands return something similar to:

  1. NAME                                READY   STATUS        RESTARTS   AGE
  2. activator-79f674fb7b-dgvss          2/2     Running       0          43s
  3. autoscaler-96dc49858-b24bm          2/2     Running       1          43s
  4. autoscaler-hpa-d887d4895-njtrb      1/1     Running       0          43s
  5. controller-6bcdd87fd6-zz9fx         1/1     Running       0          41s
  6. networking-istio-7fcd97cbf7-z2xmr   1/1     Running       0          40s
  7. webhook-747b799559-4sj6q            1/1     Running       0          41s
  1. NAME                                READY   STATUS        RESTARTS   AGE
  2. eventing-controller-69ffcc6f7d-5l7th   1/1     Running   0          83s
  3. eventing-webhook-6c56fcd86c-42dr8      1/1     Running   0          81s
  4. imc-controller-6bcf5957b5-6ccp2        1/1     Running   0          80s
  5. imc-dispatcher-f59b7c57-q9xcl          1/1     Running   0          80s
  6. sources-controller-8596684d7b-jxkmd    1/1     Running   0          83s
  1. NAME                                READY   STATUS        RESTARTS   AGE
  2. NAME                                  READY   STATUS    RESTARTS   AGE
  3. elasticsearch-logging-0               1/1     Running   0          117s
  4. elasticsearch-logging-1               1/1     Running   0          83s
  5. fluentd-ds-dqzr7                      1/1     Running   0          116s
  6. fluentd-ds-dspjc                      1/1     Running   0          116s
  7. fluentd-ds-x9xg7                      1/1     Running   0          116s
  8. grafana-59568f8f48-bz2ll              1/1     Running   0          111s
  9. kibana-logging-b5d75f556-pwzkg        1/1     Running   0          116s
  10. kube-state-metrics-5cb5c6986b-qp6pw   1/1     Running   0          116s
  11. node-exporter-bgtsb                   2/2     Running   0          112s
  12. node-exporter-cqrqv                   2/2     Running   0          112s
  13. node-exporter-xwv7f                   2/2     Running   0          112s
  14. prometheus-system-0                   1/1     Running   0          110s
  15. prometheus-system-1                   1/1     Running   0          110s

If the age of all your pods has been reset and all pods are up and running, the upgrade was completed successfully. You might notice a status of “Terminating” for the old pods as they are cleaned up.

If necessary, repeat the upgrade process until you reach your desired minor version number.