Changed
- Update values schema for generating EKS cluster.
Fleet Management
Breaking changes
kubectl gs template cluster: Either--nameor new--generated-nameparameter is now required for CAPI cluster names. We kept the CLI backward-compatible for vintage, so if none of these parameters is specified, the old default of generating a random name still applies and no error is thrown.
Fixed
- Fix team ownership
Changed
- BREAKING All values of cluster userconfig for
CAPAare moving underglobal.
- BREAKING All values of cluster userconfig for
Changed
- Changed the length of randomly-generated cluster names to 10
Added
- Add support for device authentication flow in the
logincommand and a new--device-authflag to activate it.
Fixed
- Fix storage of separate kubeconfig file (
--self-contained) for EKS workload clusters
- Add support for device authentication flow in the
⚠️ Attention: Minor release [1.2.0] contains breaking changes in default user values! The name of the service account changed to
efs-csi-safor both components. This affects clusters with already configured IRSA roles. You either need to update the trust identity policy on the role to match the new service account or set service account names to the original values via the values file. ⚠️Updated
- Extract registry from container images to allow set it for all images on the value level.
- Change node-selector values to fit the new scheme.
- Configure the same service account for both controller and node to simplify IRSA role permissions.
Changed
- Increase the cluster name length to 20 characters.
- Change how
loginworks to use our DNS record for the k8s API when using non-vintage providers, rather than the value found in the CAPI CRs. - Defaults for
cluster template --provider vspheresince it was migrated to flatcar os.
Changed
- Change default volume Docker size to 10Gb on AWS vintage NodePools.