This is a WIP repository. Basically we are using the Validated Patterns framework to deploy the resources needed for the OpenShift AI training that was done on March 4 - 6.
The goal for this pattern is to:
- Use a GitOps approach to manage the OpenShift AI training deployments across both public and private clouds.
The Validated Patterns framework also allows:
- Enable cross-cluster governance and application lifecycle management.
- Securely manage secrets across the deployment.
This basically deploys the resources found in the following repo: https://github.com/openshift-ai-examples/openshift-ai-examples/tree/main/openshift-ai-install
To deploy this do the following:
- Create an empty OpenShift environment ** You can use demo.redhat.com to deploy an OpenShift environment.
- Clone the repository
git clone https://github.com/claudiol/training-openshift-ai-deployorgit clone git@github.com:claudiol/training-openshift-ai-deploy.git - To deploy just run the following:
./pattern.sh make install
The above uses a container that includes all the tools (helm, make etc) that are needed to deploy the pattern. You can find the container definition in the following Validated Patterns repository: https://github.com/validatedpatterns/utility-container
- The Validated Patterns framework describes all the resources that will be deployed to the OpenShift cluster in the
values-hub.yamlfile.
values-hub.yaml details
The clusterGroup section defines whether this is a hub or an edge site. clusterGroup: name: hub isHubCluster: true
Understanding Namespace Creation using the Validated Patterns Framework
In the realm of Kubernetes and containerized environments, managing namespaces efficiently is pivotal. It ensures proper organization, security, and resource isolation within a cluster. With the Validated Patterns framework, creating namespaces becomes not just systematic but also highly customizable.
In this blog we will talk about the different ways to create namespaces by describing them in the Validated Patterns values files. We provide examples of the different options we support and the reasoning behind them.
Describing Namespaces in the Validated Patterns values files
Namespaces in Kubernetes offer a way to divide cluster resources between multiple users, teams, or projects. They act as virtual clusters within a physical cluster, enabling isolation and segmentation.
The Validated Patterns framework provides a structured approach to creating namespaces, allowing for a range of configurations to meet specific requirements. In addition, the Validated Patterns framework, by default, creates an Operator Group, defined by the OperatorGroup resource, which provides multi-tenant configuration to OLM-installed Operators.
The namespaces: Configuration structure
The Validated Patterns values files have several required sections that fall under the clusterGroup: top level section. The structure to describe namespaces starts with the declaration of the namespaces: section in the values-*.yaml files.
clusterGroup: name: example isHubCluster: true sharedValueFiles:
- /values/{{ .Values.global.clusterPlatform }}.yaml
- /values/{{ .Values.global.clusterVersion }}.yaml
namespaces: … The Validated Patterns framework, defined by the clusterGroup helm chart in the common github repository, accepts two formats for the namespaces: section:
- A list object
- A map object
If a namespace is described as a list item the Validated Patterns framework will use that list element and create a Namespace resource, using that name, as well as an OperatorGroup resource. By default the Validated Patterns framework will add the namespace to the spec.targetNamespaces for the OperatorGroup resource.
You can use both a list, and a map object, to describe a namespace. Here’s an example of using a list to describe a namespace.
namespaces: - open-cluster-management - vault - golang-external-secrets - config-demo - hello-world
To define labels and annotations you would describe the namespace as a map object, adding the labels and annotations needed, using the following structure in the values.yaml file:
namespaces:
- open-cluster-management: labels: openshift.io/node-selector: "" kubernetes.io/os: linux annotations: openshift.io/cluster-monitoring: "true" owner: "namespace owner"
For more information on using the Validated Patterns to create namespaces please see our blog at: The subscriptions section lists the operators that are needed by the validated pattern: https://validatedpatterns.io/blog/2023-12-15-understanding-namespaces/
Describing Operators to be deployed on OpenShift
You can also describe the subscriptions, or operators, that you need to be deployed on the OpenShift cluster. Here's an example on how to describe subscriptions in the values-hub.yaml file:
subscriptions:
- name: advanced-cluster-management
namespace: open-cluster-management
channel: release-2.3
csv: advanced-cluster-management.v2.3.2
- name: seldon-operator
namespace: manuela-ml-workspace
source: community-operators
csv: seldon-operator.v1.7.0
- name: opendatahub-operator
source: community-operators
csv: opendatahub-operator.v1.1.0
- name: openshift-pipelines-operator-rh
csv: redhat-openshift-pipelines.v1.5.1
- name: amq-streams
namespace: manuela-tst-all
channel: amq-streams-1.7.x
csv: amqstreams.v1.7.1
- name: red-hat-camel-k
namespace: manuela-data-lake-central-s3-store
channel: 1.4.x
csv: red-hat-camel-k-operator.v1.4.0
- name: red-hat-camel-k
namespace: manuela-tst-all
channel: 1.4.x
csv: red-hat-camel-k-operator.v1.4.0
Notice that the operator entry can include the target namespace where it should be installed and also the csv, or ClusterServiceVersion, of the operator. If you choose to use the CSV attribute you that will tell the operator that that is the starting version you want to use. The operator may upgrade to a higher version depending on the operator install process.
The next two sections are related to the ArgoCD environment. We describe projects and applications in the values file that will be deployed into the ArgoCD application instance.
Projects provide a logical grouping of applications, which is useful when Argo CD is used by multiple teams. An ArgoCD Application represents a deployed application instance in an environment.
projects:
- datacenter
applications:
- name: acm
namespace: open-cluster-management
project: datacenter
path: common/acm
- name: odh
namespace: opendatahub
project: datacenter
path: charts/datacenter/opendatahub
For more information on Validated Patterns go to: https://validatedpatterns.io.