Clarify that the odh namespace is only needed when the instance is being created

opendatahub-io / opendatahub.io

http://opendatahub.io/

Creative Commons Attribution Share Alike 4.0 International

2 stars 47 forks source link

Clarify that the odh namespace is only needed when the instance is being created #183

Closed adelton closed 1 year ago

adelton commented 1 year ago

Description

Clarify that the odh namespace is only needed when the instance is being created, not for the operator installation.

The order of the steps was all over the place, it also caused the odh namespace being created even if installing the 2.X operator (where per https://opendatahub.io/docs/quick-installation-new-operator/ it is not used at all).

How Has This Been Tested?

Not tested.

Merge criteria:

[x] The commits are squashed in a cohesive manner and have meaningful messages.
[ ] Testing instructions have been added in the PR body (for PRs involving changes that are not immediately obvious).
[ ] The developer has manually tested the changes and verified that the changes work

shalberd commented 1 year ago

looks good to me, but then again, I am not that close to the whole thing :-) @VaishnaviHire what do you think? This is for the "old way" of installing software via KfDef.

The new operator and its install components / CRDs mention: By default, namespace/project opendatahub is used to host all applications.

Now, does that mean the with the new operator, the namespace opendatahub must already be present or is it created by the new operator once a new DataScienceCluster CR is applied? If creating a DataScienceCluster with metadata.namespace: mycustomodhnamespace, does that namespace need to be present already or is it automatically created? In general, even with the new way of installing ODH ... the namespace name can be custom, correct?

adelton commented 1 year ago

Yes, this is solely to clarify the 1.X installation path.

With 2.1.0 that I got installed today, the opendatahub namespace/project seems to get created automatically, likely by the operator, before you even start creating the DataScienceCluster instance. At the same time, even if I put metadata.namespace to the YAML to some random value, things still got created in opendatahub. So that "by default" is really "hardcoded", it seems.

VaishnaviHire commented 1 year ago

Hi @adelton , thanks for the PR.

Yes for the v1 operator, no odh namespace is required just for the installation of the operator. The operator will be installed in openshift-operators namespace
For the v2 operator, the latest release in fast channel does create the opendatahub namespace, however we are making the namespaces configurable in upcoming versions- Ref Issue.
Note V2 operator is still in alpha release, so we will continue to actively update the code/feature sets

VaishnaviHire commented 1 year ago

looks good to me, but then again, I am not that close to the whole thing :-) @VaishnaviHire what do you think? This is for the "old way" of installing software via KfDef.

The new operator and its install components / CRDs mention: By default, namespace/project opendatahub is used to host all applications.

Now, does that mean the with the new operator, the namespace opendatahub must already be present or is it created by the new operator once a new DataScienceCluster CR is applied? If creating a DataScienceCluster with metadata.namespace: mycustomodhnamespace, does that namespace need to be present already or is it automatically created? In general, even with the new way of installing ODH ... the namespace name can be custom, correct?

As mentioned in this design doc , DataScienceCluster is a cluster scoped api, so metadata.namespace will not update applications namespace. However namespace will still be configurable by updating DSCInitialization in upcoming release

adelton commented 1 year ago

@VaishnaviHire What are the next steps to get this possibly merged?