Demystifying Operator Deployment in OpenShift

2 min read

Operators in Red Hat OpenShift clusters are the de-facto standard for adding features and capabilities of a cluster.

Applications and middleware are packaged as operators and available on the OperatorHub. Although most operators can be installed within a few clicks, some more complex operators require a deeper understanding of the infrastructure. Similar to the water for a kitchen sink, most people just need to know that it is available; however, knowing the plumbing underneath the surface is necessary for problem-solving and fixing errors when things do not work as expected.

This article attempts to explain the underlying objects and processes that make up the operators and operator framework. The content in this article is divided into extending the OperatorHub and the deployment of an Operator.

Extending the OperatorHub

The OperatorHub is populated from the content in OperatorSource and CatalogSources. Most of the newer sources are now using the CatalogSource format. I will explain the difference between the CatalogSource and OperatorSource and how they work in a future article:

OperatorSource and CatalogSource.

OperatorSource and CatalogSource.

You can view these sources from the Web console under Administration > Cluster Settings > Global Configuration > OperatorHub > Sources. The following is a screenshot of this menu:

OperatorHub sources.

OperatorHub sources.

The catalog source consists of a non-executable container image. The container image contains a file that acts as a catalog of PackageManifests that can be installed. When a CatalogSource is defined, OpenShift creates a Job to load the catalog image, retrieve the individual PackageManifest and create the objects in OpenShift. Each PackageManifest object is a tile that you can see in the Operators > OperatorHub menu of the OpenShift Web Console:

CatalogSource and PackageManifest

CatalogSource and PackageManifest.

Each of the PackageManifest objects contains a unique definition on how to implement the operators, including the following:

  • Channels: The path for installation and upgrade of an operator package.
  • Cluster Service Version: Package definition for a certain version of the operator, the CSVs allow the operator that subscribes to a channel to dynamically evolve (upgrade).
  • Custom Resource Definition: Part of the CSV that defines the structure of a Custom Resource that the Operator will be managing. 
  • Container images: Images that will be loaded when you install this CSV.

Operator deployment

When you choose to install an Operator from OperatorHub, you create a Subscription object. It is subscribing to a channel in the PackageManifest. The notion of subscribing allows an automatic update (as defined in the installPlanApproval field) when the CSV in the PackageManifest is updated:

Channel and Subscription

Channel and Subscription.

The CSV from the channel is built and generates an installPlan, which contains a list of resources that should be created for this operator. Subscription also defines the Custom Resource Definition that is managed by this operator. Once the installation is successful (the CSV phase becomes Succeeded from the oc get csv command), that indicates that the Operator is installed:

Installed operator

Installed operator.

Once an operator is installed, you have a Deployment with a pod that runs the operator controller process. The operator controller runs a loop that monitors the Custom Resources in its namespace (or all namespaces as defined by the installation method). As a Custom Resource is created, it may perform additional tasks, such as creating more resources in the cluster.

The illustration above triggers the creation of the OpenShift Container Storage cluster based on the content of the StorageCluster custom resource.

Learn more about IBM Garage.

Be the first to hear about news, product updates, and innovation from IBM Cloud