feat(cloud-provider): add single node known issue

[FrankYang0529]

issue: harvester/harvester#6799

[github-actions]

Name	Link
🔨 Latest commit	4b05aca
😎 Deploy Preview	https://6732ca6989a1ed25dbc9e09f--harvester-preview.netlify.app

[starbops]

LGTM, thanks.

[jillian-maroket]

@asettle The new content is for Prime customers. How do you want to handle this?

[asettle]

@asettle The new content is for Prime customers. How do you want to handle this?

This release note should only go into prime docs. As a community user, you shouldn't be worried about upgrading from community to Prime. However, Prime users will be as they are already looking to purchase and migrate from Prime. This should be kept separate and what we want to be saying is, "When upgrading your Rancher version from the community version to the SUSE supported Prime version, the harvester-cloud-provider pod won't be ready in a single-node cluster."

[jillian-maroket]

@FrankYang0529 Below is the edited version. Please check if the technical details are correct. I will add the content to the correct repo tomorrow.

You can upgrade Rancher from the community version to the Rancher Prime version in a single-node SUSE Virtualization cluster. However, the harvester-cloud-provider pod will not be ready after the upgrade is completed.

The default container registry used for images (system-default-registry) changes during the upgrade. All pods in the guest cluster receive the updated value for system-default-registry, but the harvester-cloud-provider pod (in single-node clusters) is unable to reach the ready state. This occurs because the rolling update strategy (.spec.strategy.rollingUpdate.maxUnavailable) is set to 25%, so the pod deployment update becomes stuck.

The issue was fixed in Rancher 2.8.x (103.0.3+up0.2.6) and Rancher 2.9.x (104.0.2+up0.2.6). In earlier versions, the workaround is to manually remove the old harvester-cloud-provider pod.

cc: @bk201

[FrankYang0529]

Hi @jillian-maroket, the content LGTM. Thanks for updating it 👍 .