[Feature] Support private GKE cluster on Google Cloud

[zijianjoy]

This is a survey form for collecting interest on the feature: Private Kubeflow Cluster

Description

GKE allows you to create private GKE cluster http://cloud/kubernetes-engine/docs/how-to/private-clusters which doesn't expose external IP address to the public. It requires customization and deployment steps change in kubeflow/gcp-blueprints.

Action

If this feature fits in to your use case, please upvote this GitHub issue. We will prioritize this work accordingly, also welcome contributor from the community.

[kungfuchicken]

tracking.

[markcberman]

Are there any instructions for deploying latest stable KF version to a private GKE cluster. My company's firewall rules block access to the public internet.

[4j4y]

Any update on this?

[bsikander]

I think we should clarify here and in the documentation that the managed version of kubeflow pipelines (AI Platform Pipelines) on GCP, do work with private GKE cluster. Only custom deployment of kubeflow on private GKE requires manual steps. For new comers, this is totally not clear.

[zijianjoy]

@bsikander That is a good point. However, AI Platform Pipelines is the managed version of Kubeflow Pipelines standalone, not Full fledged Kubeflow. Therefore AI Platform Pipelines guidance lives in https://www.kubeflow.org/docs/components/pipelines/installation/overview/#google-cloud-ai-platform-pipelines. Feel free to contribute documentation there to explain that AI Platform Pipelines works with private GKE cluster. The Distribution section is more about Full fledged Kubeflow instructions.

[bhack]

Any update on this?