First of all, I appreciate your work on this project and thank you for your
I'm currently trying to use a mix of ODH/Kubeflow tools to cover a complete
AI infrastructure. I'm still a little bit confused about which components
to use :
- airflow vs kubeflow pipelines
- seldon core vs kubeflow kfserving
- kubeflow's jupyterlab vs odh's jupyterhub notebooks
To help me make my decision, I'd like to know which are the components that
you guys think are stable enough.
For example, I think the airflow operator is still in v1alpha version and
the UI is not yet accessible from odh dashboard and thus not yet suited for
production, is it?
If you can guide me to where I can find useful information I'd appreciate
Thanks in advance.
I would like to know if there is any way to achieve multi-tenancy in
kubeflow's openshift distribution?
I'm on openshift 4.7 and I did manage to install kubeflow using the
Once installed, there were no authentication mechanism and thus not ideal
for production use-cases.
I did some search and figured out opendatahub's teams are reflecting on an
integrated authentication solution for ODH and KF (cluster vs namespace).
Meanwhile, I still need multi-tenancy and I would like to know if it's
possible to have kubeflow installed with istio and dex(or keycloak) on
Thanks in advance.
We are happy to announce a new z-stream release for Open Data Hub v1.1.1 In
this release we migrated the apiVersion of the operator to v1 so it is
visible in OperatorHub on OCP 4.9. We have provided an example kfdef
installing ODH on OCP 4.9 with limited components. The team is working on
getting the rest of the components working on OCP 4.9, to follow this
effort please check our jira epic <https://issues.redhat.com/browse/ODH-522>
The ODH v1.1.1 release also includes many new features described below in
our operator description.
- JupyterHub v0.3.5 - Open source multi-user JupyterLab notebook
platform w/ GPU support and *NEW* High Availability.
- Trino v355 - Distributed analytics SQL database that supports multiple
- Hue v4.8.0 - A service that provides data exploration on Hive and S3
- Spark Thrift Server - A service that allows JDBC clients run Spark SQL
- Open Data Hub Dashboard v1.0 - A web dashboard that displays installed
Open Data Hub components with easy access to component UIs and documentation
- Elyra v2.2.4 - JupyterLab notebooks with support for AI workflows
- Ceph Nano v0.7 - Minimal Object Storage provided by Ceph for
- Apache Spark v2.4.5 - Unified analytics engine for large-scale data
- Prometheus v2.16.0 - Monitoring and alerting tool
- Grafana v7.1.1 - Data visualization and monitoring
- Airflow v1.10.11 - Workflow management
- Seldon v1.2.0 - Open source platform for deploying machine learning
- Argo v2.12.5 - Container-native Workflow Engine
- Apache Superset v1.3.0 - *NEW* Open source application for data
exploration and visualization
- Apache Kafka v2.8.0 - The open source stream processing platform
- OpenShift Pipelines v1.3.1 - Cloud-native CI/CD on OpenShift To
install one or multiple of these components use the default KfDef provided
with the operator.
ODH 1.1.1 supports Kubeflow v1.3.0 and some components such as KF Serving
and KFP on Tekton from master branch. To install Kubeflow v1.3.0 components
please use this example KfDef
NOTE: The Open Data Hub distribution of the Kubeflow 1.3 stack currently
runs on OpenShift versions of 4.8 (or lower). For latest updates, please
visit link <https://github.com/opendatahub-io>
For installing Open Data Hub on OpenShift 4.9, please use the provided
only includes JupyterHub. We are currently updating all other components to
work on OpenShift 4.9 for the latest update please visit ODH-522