Kubeflow on Google Cloud - 1.8.0 release tracker

[Linchin]

Bugs/Issues to investigate:

• Test Kubeflow 1.8.0 on GCP
  • deployment
  • test pipeline run
  • test notebook
  • test KFP SDK 2.4
• Test Kubeflow 1.8.0-rc.5 on GCP
  • deployment
  • test pipeline run
  • test notebook
  • test KFP SDK 2.4
• Test Kubeflow 1.8.0-rc.2 on GCP
  • deployment
  • test pipeline run
  • test notebook
• Test Kubeflow 1.8.0-rc.1 on GCP
  • deployment
  • test pipeline run
  • test notebook
• Test Kubeflow 1.8.0-rc.0 on GCP
  • deployment
  • test pipeline run
  • test notebook
• kubeflow.org_profiles.yaml causes errors with creationTimestamp line #431
• kustomize generates file with creationTimestamp line #440
• [sdk] SDK client cannot connect to Kubeflow on GCP deployment from a Notebook environment. kubeflow/pipelines#10075

Features to consider (carry over from last release):
Out of scope:

• Fix when GKE reaches version 1.27 GKE Default prometheus enable feature causing conflict #422
• Upgrade to Kustomize 5.0: fixes On installation Kustomize v5.0.0 decode error #402
• set up automated testing #437
• Explore using Porch from kpt: https://kpt.dev/guides/porch-installation
• Consider using profile controller to add labels for user created namespaces, then we can use label to distinguish user workload.
• Migrate from Config Controller to Config Delivery
• Explore upgrading controller tool version
• Explore using Terraform to enable GCP services instead of using management cluster.

/CC @chensun @zijianjoy @gkcalat

[chensun]

Not sure how we marked notebook as tested. It seems there's a authentication issue from the SDK implementation that would block any SDK interaction with the pipeline from a Notebook environment.
kubeflow/pipelines#10075