Attention Users using a Private Registry for storing the Kubernetes component images on Rancher Launched Kubernetes/RKE clusters
Do not use Rancher v2.2.5 or use RKE CLI v0.2.5
Note: This does not affect any workloads using private registries. This is specific to the private registry used for pulling the Kubernetes component images.
When using a private registry to pull the Kubernetes component images for a Rancher Launched Kubernetes cluster within Rancher or with the RKE CLI, etcd backups/snapshots fail to succeed because the image being used to create the backup/snapshot will not be pulled from the configured registry.
Here are the use cases where you might hit this issue:
• Air-gapped Rancher setups - You are in an air-gapped Rancher setup and you have configured the
system-default-registry within Rancher. This could have been done either in the UI of Rancher or passed as an environment variable (
CATTLE_SYSTEM_DEFAULT_REGISTRY) when starting the Rancher server container. When creating any Rancher Launched Kubernetes cluster, you will hit this issue in Rancher v2.2.5.
• For any Rancher Launched Kubernetes cluster that has enabled a private registry - If you are either using node pools or custom clusters and you have selected to use a private registry for pulling images for launching the cluster, you will hit this issue in Rancher v2.2.5.
• RKE CLI users using the private registry directive - You have specified a private registry for your system images in your cluster.yml. Using RKE CLI v0.2.5, if you run
rke up, you will hit this issue.
This issue is being tracked in GH #1472. If you have already encountered this issue, there is a workaround, but we are aiming to ship Rancher v2.2.6 and RKE v0.2.6 as soon as possible to fix this issue.