Upgrade

A Verrazzano installation consists of a stack of components, such as cert-manager, where each component has a specific release version that may be different from the overall Verrazzano version. The Verrazzano platform operator knows the versions of each component associated with the Verrazzano version. When you perform the initial Verrazzano installation, the appropriate version of each component is installed by the platform operator. Post installation, it may be necessary to update one or more of the component images or Helm charts. This update is also handled by the platform operator and is called an upgrade. Currently, Verrazzano does only patch-level upgrades, where a helm upgrade command can be issued for the component. Typically, patch-level upgrades simply replace component images with newer versions.

Application and system pod restarts

If Verrazzano has a new version of Istio, then all the pods with Istio proxy sidecars need to be restarted. This is done so that the new version of the proxy sidecar can be injected into the pods. All Verrazzano pods containing Istio proxy sidecars will be restarted. This includes Verrazzano system pods, such as the NGINX Ingress Controller, along with Verrazzano applications. For WebLogic workloads, Verrazzano will shut down every domain, do the upgrade, then restart every domain. For all other workloads, Verrazzano will perform a rolling restart when the upgrade has completed. There is no user involvement related to restarting applications; it is done automatically during upgrade.

Upgrade steps

It is important to distinguish between updating the Verrazzano platform operator versus upgrading the Verrazzano installation. The platform operator contains the newer component charts and image versions, so it must be updated prior to upgrading the installation. Updating the platform operator has no effect on an existing installation until you initiate the Verrazzano installation upgrade. Currently, there is no way to roll back either the platform operator update or the Verrazzano installation upgrade.

You can upgrade Verrazzano using the Verrazzano CLI or with kubectl. See the following respective sections.

$ kubectl patch vz example-verrazzano -p '{"spec":{"version":"v1.4.8"}}' --type=merge

Verify the upgrade

Check that all the pods in the verrazzano-system namespace are in the Running state. While the upgrade is in progress, you may see some pods terminating and restarting as newer versions of components are applied, for example:

$ kubectl get pods -n verrazzano-system

# Sample output
coherence-operator-866798c99d-r69xt                1/1     Running   1          43m
fluentd-f9fbv                                      2/2     Running   0          38m
fluentd-n79c4                                      2/2     Running   0          38m
fluentd-xslzw                                      2/2     Running   0          38m
oam-kubernetes-runtime-56cdb56c98-wn2mb            1/1     Running   0          43m
verrazzano-application-operator-7c95ddd5b5-7xzmn   1/1     Running   0          42m
verrazzano-authproxy-594d8c8dcd-llmlr              2/2     Running   0          38m
verrazzano-console-74dbf97fdf-zxvvn                2/2     Running   0          38m
verrazzano-monitoring-operator-6fcf8484fd-gfkhs    1/1     Running   0          38m
verrazzano-operator-66c8566f95-8lbs6               1/1     Running   0          38m
vmi-system-es-master-0                             2/2     Running   0          38m
vmi-system-grafana-799d79648d-wsdp4                2/2     Running   0          38m
vmi-system-kiali-574c6dd94d-f49jv                  2/2     Running   0          41m
vmi-system-kibana-77f8d998f4-zzvqr                 2/2     Running   0          38m
weblogic-operator-7b447fdb47-wlw64                 2/2     Running   0          42m

Check that the pods in your application namespaces are ready, for example:

$ kubectl get pods -n todo-list

# Sample output
NAME                     READY   STATUS    RESTARTS   AGE
mysql-67575d8954-d4vkm   2/2     Running   0          39h
tododomain-adminserver   4/4     Running   0          39h

Upgrade failures

In Verrazzano 1.3 and later, upgrade will continue to run until it succeeds or until you delete the Verrazzano CR. In previous versions, upgrade could fail and transition to the UpgradeFailed state. If that happens, and you updated the Verrazzano platform operator to 1.3+, then the Verrazzano CR will transition to UpgradePaused. To continue with the upgrade, you must change the CR version to the current version of the Verrazzano platform operator. The following steps illustrate this scenario:

1. You install Verrazzano 1.1.2.
2. You upgrade to 1.2.0 by changing the Verrazzano CR version field to v1.2.0.
   • For some reason, the upgrade failed and the Verrazzano CR state transitions to UpgradeFailed.
3. You update the Verrazzano platform operator to 1.3.0.
   • The Verrazzano CR state transitions to UpgradePaused.
4. You change the Verrazzano CR version field to v1.3.0.
   • The Verrazzano CR state transitions to Upgrading and stays in that state until it completes, then it transitions to UpgradeComplete.

To see detailed progress of the upgrade, view the logs with the following command.

$ kubectl logs -n verrazzano-install \
    -f $(kubectl get pod \
    -n verrazzano-install \
    -l app=verrazzano-platform-operator \
    -o jsonpath="{.items[0].metadata.name}") | grep '^{.*}$' \
    | jq -r '."@timestamp" as $timestamp | "\($timestamp) \(.level) \(.message)"'