Current Series Release Notes¶

1.2.0¶

Default helm repository is updated to ‘azimuth-cloud.github.io/capi-helm-charts’.

Helm repository ‘stackhpc.github.io/capi-helm-charts’ is deprecated. The last version supported on this repo is 0.9.0

Fixes an issue where creating a cluster with the 2 or more nodes in the default worker group and with auto_scaling_enabled=true but with neither {min,max}_node_count labels set would lead to a situation where the default worker group would only have 1 node instead of 2.

This is the initial release of a new in-tree Magnum driver which leverages Helm and the Kubernetes Cluster API project to provide an alternative to the existing Heat driver for creating and managing Kubernetes clusters via OpenStack Magnum.
This project started life as a out-of-tree driver which was developed and maintained at https://github.com/stackhpc/magnum-capi-helm. This 1.0.0 release marks the deprecation of this out-of-tree source repository and the first official release of the in-tree version which is under OpenStack governance, with source code now available at https://opendev.org/openstack/magnum-capi-helm.
The driver is generally ready for production use and has been deployed on a number of OpenStack systems prior to this first official release. However, this does not preclude the possibility of unforeseen issues and omissions at this early stage of the release cycle - bug reports welcome and appreciated!

Adds support for autoscaling of non-default worker node groups. The min/max node count properties on each node group are passed to the autoscaler to determine autoscaling behaviour. If the target cluster has the auto_scaling_enabled cluster label set to true then any node groups with both min_node_count` and max_node_count properties set will be allowed to autoscale between these min and max node counts depending on the cluster’s current workload.

Adds support for autoscaling of the default worker node group by respecting the min_node_count and max_node_count cluster labels. Unlike the Heat driver, when these labels are provided, the cluster will always start with node count equal to the minimum node count value and the alternative node_count flag will be ignored.