Skip to content

Latest commit

 

History

History
75 lines (57 loc) · 3.07 KB

README.md

File metadata and controls

75 lines (57 loc) · 3.07 KB

MultiCluster Engine Operator

Multicluster engine operator installs and manages multicluster components, such as cluster-manager, hive, and other core foundation resources.

Prereqs

You must meet the following requirements to install the MultiCluster Engine:

  • An OpenShift Container Platform (OCP) 4.3+ cluster available
    • You must have a default storage class defined
  • oc (ver. 4.3+) & kubectl (ver. 1.16+) configured to connect to your OCP cluster
  • oc is connected with adequate permissions to create new namespaces in your OCP cluster.

Getting Started

To install the MultiCluster Engine, call the helper script which will supervise the installation of multicluster components. This script requires a SNAPSHOT tag as input.

  1. Follow these presteps if deploying MultiCluster Engine downstream builds:

    1. Follow steps here to set up ImageContentSourcePolicy and configure the global pull secret.
    2. Set the following environment variable -
    export DOWNSTREAM=true
  2. Run the ./multiclusterengine/start.sh script in the root directory of this repository

$ ./multiclusterengine/start.sh
  1. When prompted for the SNAPSHOT tag, either press Enter to use the previous tag, or provide a new SNAPSHOT tag.

After the tag has been provided, the installation will continue. Currently the installation deploys and manages its components in the multicluster-engine namespace which it creates.

Upgrading (Development only)

You can upgrade you UPSTREAM Multicluster Engine with the upgade.sh script.

Run interactive

  1. Run the script
    # Use the optional --watch parameter
    ./upgrade.sh
  2. You will be prompted to enter the snapshot, make sure you choose a 2.0.0-BACKPLANE-... snapshot
  3. If you added the --watch parameter, a watch on the pods will start, so you can see the changes.
    • You will ALWAYS see the multicluster-engine-operator-... pods (two of them) restart.
    • Any other images that are new to the snapshot will be restarted
    • The pods are sorted with NEWEST last
    • When the multicluster-engine-operator restarts, it will apply any CRD changes
  4. Press ctrl-c to end the watch!

Run with no input

  1. Run the script, including the snapshot
    export snapshot=2.0.0-BACKPLANE-2022-04-29-01-53-52 && ./upgrade.sh
  2. The upgrade is performed, but the script does not WAIT for any pod change to occur

Uninstallation

To uninstall the MultiCluster Engine, follow these steps:

  1. Delete the multiclusterengine custom resource -
$ oc delete multiclusterengine --all
  1. Delete the MultiCluster Engine CSV, Subscription, and namespace.
$ oc delete csv $(oc get sub multicluster-engine -o jsonpath='{.status.currentCSV}')
$ oc delete sub multicluster-engine
$ oc delete namespace multicluster-engine