Managing Operators with OLM
OLM is the second pillar of the Operator Framework. Its purpose is to facilitate the deployment and management of Operators in a Kubernetes cluster. It is a component that runs within a Kubernetes cluster and provides several commands and features for interacting with Operators.
OLM is primarily used for the installation and upgrade of Operators—this includes fetching and installing any dependencies for those Operators. Users interact with OLM via commands provided by the Operator SDK binary, the Kubernetes command-line tool (kubectl), and declarative YAML.
To get started, OLM can be initialized in a cluster with the following command:
$ operator-sdk olm install
Besides installing Operators, OLM can also make Operators that are currently installed discoverable to users on the cluster. This provides a catalog of already installed Operators available to cluster users. Also, by managing all the known Operators in the cluster, OLM can watch for conflicting Operator APIs and settings that would destabilize the cluster.
Once an Operator's Go code is compiled into an image, it is ready to be installed into a cluster with OLM running. Technically, OLM is not required to run an Operator in any cluster. For example, it is completely possible to deploy an Operator manually in the cluster, just as with any other container-based application. However, due to the advantages and security measures described previously (including its ability to install Operators and its awareness of other installed Operators), it is highly recommended to use OLM to manage cluster Operators.
When developing an Operator, the image is compiled into a bundle, and that bundle is installed via OLM. The bundle consists of several YAML files that describe the Operator, its CRD, and its dependencies. OLM knows how to process this bundle in its standardized format to properly manage the Operator in a cluster.
Compiling an Operator's code and deploying it can be done with commands such as the ones shown next. The first command shown in the following code snippet builds the bundle of YAML manifests that describe the Operator. Then, it passes that information to OLM to run the Operator in your cluster:
$ make bundle ... $ operator-sdk run bundle ...
Later chapters will demonstrate exactly how to use these commands and what they do, but the general idea is that these commands first compile the Operator's Go code into an image and a deployable format that's understandable by OLM. But OLM isn't the only part of the Operator Framework that consumes an Operator's bundle—much of the same information is used by OperatorHub to provide information on an Operator.
Once an Operator has been compiled into its image, OperatorHub exists as a platform to share and distribute those images to other users.