Kubernetes Enhancements¶

Helm Installs¶

Helm is now required to do Kubernetes installs and is used by the PSO installer to deploy to the cluster. The installer now takes YAML as input for Kubernetes installs rather than XML. Multiple YAML files can be passed to the installer and they will be merged together.

The YAML input file includes all the data needed for the database setup part of the installation and also the Helm deployment part of the installation. A Role Configuration file is no longer provided as input and instead the number of instances and resources required for each application is specified in the deployments section of the YAML to override the defaults.

See the Installation Guide for further details on using YAML based Helm installations.

Change of URLs¶

The default URLs for accessing the applications have changed from /automation/gateway and /automation/workbench to /pso/gateway and /pso/workbench. Upgraded existing installs will continue to use the /automation/* URLs. This can be configured by specifying a different baseHref under the deployment yaml section for the relevant applications.

Change of container image names¶

The container image names have changed from ifs-automation: and ifs-automation-python: to ifs-pso: and ifs-pso-python:. When pulling the images to host in a private registry both must be pulled and hosted using the new name.

Network policies¶

Network policies are now installed into the cluster to limit network interaction between the installed pods and other pods in the cluster. Network policies require a network plugin to be active in the cluster to apply. See the Kubernetes documentation for more information. Egress rules may need to be provided in the YAML file to allow communication from the pods to the database or external services. The network policies are enabled by default but can be disabled.

Auto-Scaling¶

Horizontal pod auto-scaling is now supported via the use of KEDA. Auto-scaling is disabled by default but can be enabled at a global level or per application. KEDA must be installed into the cluster for this to work and the Kubernetes Metrics Server must also be available to allow KEDA to scale based on CPU and Memory.

The method of scaling is different for each application and is based on metrics such as the CPU, Memory or number of active datasets the system. The DSE also supports scaling down to zero when there is nothing to process in the system. The details of this can be found in the Administration Guide.

Minor Enhancements¶

The WCF Gateway can now be installed into Kubernetes clusters to support existing integrations but is disabled by default. The WCF Gateway continues to be deprecated with the use of the RESTful Gateway advised for new integrations.

See the Installation Guide for more information.