Running operators in their own namespace

[fgiloux]

Currently operators (OpenShift Pipelines for now) are installed in openshift-operators namespace.

There are benefits of installing the operators in their own namespaces (still separated from operands) rather than openshift-operators. It would however require an OperatorGroup first to be created.

• better segregation
• dedicated InstallPlan, which provides for instance the possibility to have different upgrade strategy by operator. This is really useful when one runs into troubles.
• operartors with different scopes don't cohabit very well in the same namespace

It is not an issue as long as we don't install other operators but things may change.

[adambkaplan]

It is not an issue as long as we don't install other operators but things may change.

There's a good chance that we will install other operators in our base deployment on workload clusters. The first to come to mind are the ElasticSearch operator and the OpenShift Logging operator: https://docs.openshift.com/container-platform/4.10/logging/cluster-logging.html#cluster-logging-about_cluster-logging.

These come to mind as our "first option" for forwarding TaskRun logs.

[fgiloux]

We will be fine with these ones as they get installed by default under the following namespaces:

• openshift-operators-redhat
• openshift-logging