Update dependency kubernetes-sigs/cluster-api-provider-openstack to v0.10.3 (!2309) · Merge requests · Sylva-projects / sylva-core

Renovate Bot requested to merge renovate/kubernetes-sigs-cluster-api-provider-openstack-0.x into main May 23, 2024

This MR contains the following updates:

Package	Update	Change
kubernetes-sigs/cluster-api-provider-openstack	minor	`v0.9.0` -> `v0.10.3`

Release Notes

kubernetes-sigs/cluster-api-provider-openstack (kubernetes-sigs/cluster-api-provider-openstack)

Changes since v0.10.2

🐛 Bug Fixes

Handle errors returned by GetInstanceStatusByName in machine controller (#2087)
allNodesSecurityGroupRules: relax remote fields (#2080)
Fix loadbalancer timeout panic (#2076)
Fix empty version output in release builds (#2059)
Fix panic executing manager without valid kube context (#2061)
Fix nil pointer issue while creating port (#2065)

🌱 Others

Drop dulek from reviewers (#2083)
Set FallbackToLogsOnError on CAPO manager (#2072)
Refactoring: never assign unacceptable TLS versions (#2062)

Thanks to all our contributors! 😊

Breaking API Changes

v0.10.0 is a major update which brings major changes to the API.

v1alpha5 is no longer served

If you are still using v1alpha5, this will not work in v0.10.0. However, for this release only objects are still defined in the CRDs and the code is still present, so as a temporary workaround it is possible to manually edit the CRDs to set versions.served to true for v1alpha5 objects. This is not tested, and we have low confidence that this will work without problems. Some manual effort may be required to check and fix automatically converted objects.

v1alpha6 and v1alpha7 are deprecated

v1alpha6 and v1alpha7 objects will be automatically converted to v1beta during use. This is well tested. We don’t anticipate problems with these conversions.

We will stop serving and testing v1alpha6 in the next release.

v1alpha7 is not marked deprecated in v0.10.0 to allow a switch-over period without deprecation warnings, but will be marked deprecated in the next release. Will will stop serving and testing it in a release after that.

You should update to use v1beta1 natively as soon as possible.

v1beta1 is released

v1beta1 marks a major update to the CAPO API. The specific changes from v1alpha7 are documented here: https://cluster-api-openstack.sigs.k8s.io/topics/crd-changes/v1alpha7-to-v1beta1

More than this, though, it marks an intention by the maintainers to stop making breaking changes. The API will continue to evolve, but we will make every effort to do this without introducing more backwards-incompatible changes.

Removal of hardcoded Calico CNI security group rules

This is documented more completely in the API upgrade documentation.

Prior to v1beta1, when using managed security groups we would automatically add certain rules which were specific to Calico CNI. It was not possible to add rules for any other CNI. A common way to work round this was to set allowAllInClusterTraffic: true.

With v1beta1 there are no longer any implicit rules for any CNI. However, it is now possible to specify custom rules in the cluster spec which will be automatically added to managed security groups. Users of Calico CNI must now add these rules explicitly. Users of other CNIs now have the option of using managed security groups.

Calico CNI rules will be added automatically when upgrading to v1beta1 from a previous API version.

The Calico CNI rules have been added to the release templates, so for now creating a cluster with clusterctl will continue to have Calico rules when using the default templates.

Management cluster changes

Removal of MutatingWebhookConfiguration

CAPO no longer uses a mutating webhook, and its configuration is removed. If you upgrade your management cluster with clusterctl this will be handled correctly. If you do it manually you must ensure you remove the MutatingWebhookConfiguration capo-mutating-webhook-configuration. If you do not you may see errors like the one in https://github.com/kubernetes-sigs/cluster-api-provider-openstack/issues/1927.

Minimum management cluster version is now 1.25

v0.10.0 now uses https://kubernetes.io/docs/reference/using-api/cel/ for some API validations, which only became available without a feature gate in 1.25. Consequently we now require the management cluster to be at least k8s 1.25.

Highlighted new features

API Reference documentation

We now automatically publish API reference documentation! The documentation for v1beta1 can be found here: https://cluster-api-openstack.sigs.k8s.io/api/v1beta1/api

Floating IP IPAM Provider

It is now possible to allocate floating IPs for individual machines using the new Floating IP IPAM Provider documented here: https://cluster-api-openstack.sigs.k8s.io/api/v1alpha1/api#infrastructure.cluster.x-k8s.io/v1alpha1.OpenStackFloatingIPPool

Attach them to a machine via the new floatingIPPoolRef in OpenStackMachineSpec: https://cluster-api-openstack.sigs.k8s.io/api/v1beta1/api#infrastructure.cluster.x-k8s.io/v1beta1.OpenStackMachineSpec