Just a quick update on how I'm currently (manually) testing helm upgrade...

1. Create kind cluster
2. make in 2.2.x branch for CRDs and 2.2.1 images
3. kind load docker-image the just-built 2.2.1 operator & DAC images
4. helm install an older operator version (using --version 2.0.209 in this case)
5. Wait for it to start up, pods running, etc.
6. kubectl replace and then kubectl create the new (2.2.1) crd.yaml
7. helm upgrade [name] /path/to/helm/charts/couchbase-operator (using Tommie’s fork as it’s currently pointing at 2.2.1) - also using --values so we’re using the locally build 2.2.1 files instead of the (currently non-existent) couchbase/operator:2.2.1

There is an extra step when upgrading from 2.0.2 (to anything >= 2.1) on helm, to do with Secrets. Tommie explains it in a comment here. Adding cluster.security.adminSecret=<release-name>-couchbase-cluster resolves this.

Manually tested helm upgrading 2.0.2 (no 2.0.3 helm chart), 2.1.0, and 2.2.0 to 2.2.1, all worked fine. Automation for this to be added in 2.3.

Re-did these tests on a new Helm PR from Patrick, which was slightly different from Tommies (which I initially tested on) - worked fine just the same