Comprehensive ease of use: OpenClusterManagement 0.7 released

Author: Zuo Xiu

OCM version 0.7 released

After a long wait, the OpenClusterManagement community officially released the v0.7 version at the end of March. In the new version, there are a series of new functions and features to welcome interested readers to experience and explore. At the same time, in this version, the community maintainers have also fixed some problems with the existing functions and polished and improved the experience for end users. Newly logged in users can refer to document [ 1] for experience, and users who have deployed version 0.6 can also refer to document [ 2] to upgrade the existing environment.

New feature DefaultClusterSet

In order to enable users to smoothly log in to OCM's cross-cluster orchestration capabilities (such as Placement), a new feature "DefaultClusterSet" is supported in the new 0.7 version. In previous OCM versions, users needed to plan the ClusterSet topology of the cluster in OCM in advance to experience related advanced features. Now, all managed clusters registered in the OCM environment will be registered with the name "default" by default. in the ClusterSet. This allows us to perform multi-cluster orchestration in a multi-cluster environment with a relatively simple topology by directly binding/mapping "default" to the namespace of a central cluster.

Please refer to document [ 3 ] on how to plan multi-cluster topology based on ClusterSet in OCM.

title=

Placement API evolves to v1beta1 version

After the community feedback on the v1alpha1 version, the OCM community officially evolved the Placement API to the v1beta1 version. The Beta version means that the community will provide a more reliable guarantee for the compatibility maintenance of the API model of this version. At the same time, in order to simplify the experience of the Placement API for end users, the Placement API will support multi-cluster scheduling based on Taint/Toleration semantics in the beta version. In this way, users who are familiar with the native scheduling mechanism of Kubernetes can refer to the pattern of Tainting nodes for a single cluster/Toleration for containers and apply it to a multi-cluster scenario: we can tag some clusters with Taints and declare them in the Placement API. Toleration has implemented dynamic multi-cluster scheduling.

title=

Hub cluster version requirement reduced from 1.19 to 1.12

In previous OCM versions, the requirement for the Hub cluster version was 1.19 or higher, mainly because the OCM hub components depended on the CSR API of the GA version to work. Compatibility with the Beta version of the CSR API is now available in version 0.7, but currently this compatibility needs to be enabled manually by adding the following configuration parameters to the registration component of OCM:

--feature-gate=V1beta1CSRAPICompatibility=true

Hosted Deployment Mode

The default deployment mode of OCM is "hub-spoke" [ 4 ] mode, that is, one or more Agent controllers are deployed in each managed cluster to operate the cluster, which is also the "Pull" architecture mode often referred to in multi-cluster. In the new version, OCM supports adjusting the deployment structure to the deployment mode in which the Agent controller moves up and refers to the central cluster, which we call "Hosted deployment". In Hosted mode, no other components need to be deployed in the hosting cluster, and all proxy controllers are executed remotely.

OCM and KubeVela 1.3 release enhanced multi-cluster feature integration

The KubeVela 1.3 version released at the same time further integrates OCM and KubeVela. You can refer to the above operation guide and screen recording for experience. In the above guide, we can experience:

How to deploy OCM hub component Hub for multi-cluster environment through KubeVela's plug-in mechanism
How to deploy the OCM agent controller component Klusterlet through the vela command line for managed clusters
Experience KubeVela 1.3 multi-cluster application publishing function

Alibaba Cloud ACK Agile Edition OCM Practice

In the new version of Alibaba Cloud ACK Agile Edition [ 6] , the multi-cluster proxy gateway plug-in that has logged in OCM is fully integrated. We can also quickly experience in our own OCM environment according to the document [ 7] . In general, through the multi-cluster proxy gateway plug-in, we can enable the components in the OCM central cluster to access the control plane of the managed cluster across any network infrastructure, and the client keys for accessing the managed cluster will also be dynamically rolled to avoid copy leakage. And so on security issues. We can even register any KinD cluster on the local laptop with the OCM hub on the cloud and have forward API access.

Reference link:

[1] Documentation

https://open-cluster-management.io/getting-started/quick-start/

[2] Documentation

https://open-cluster-management.io/getting-started/administration/upgrading/

[3] Documentation

https://open-cluster-management.io/concepts/managedclusterset/

[4] hub-spoke

https://en.wikipedia.org/wiki/Spoke%E2%80%93hub_distribution_paradigm

[5] OCM and KubeVela Version 1.3 Enhanced Multi-Cluster Function Integration Operation Guide**

https://github.com/oam-dev/samples/tree/master/12.Open_Cluster_Management_Demo

[6] Alibaba Cloud ACK Agile Edition