MiniKF 20220916.0.0

September 21, 2022

New MiniKF release including:

  • Kubeflow 1.4
  • Rok 1.5.3-28-g610fd714c

New Features

  • Upgrade Rok to version 1.5.3.
  • Upgrade MiniKF to Ubuntu Focal 20.04.
  • Secure and encrypt network traffic for Kubeflow Pipelines and KFServing by using Istio sidecars with mTLS.
  • Get full control over the deploy configuration of your Kale pipelines, both from the Kale UI and the Kale SDK.
  • Access a monitoring dashboard to get a holistic view of your Kubernetes and Rok cluster.
  • Write a simple Python for-loop and let Kale convert it to hundreds of pipeline steps without having to write any KFP DSL or produce any Docker images.
  • Call step functions with keyword arguments in the Kale SDK.
  • Produce pipeline components that are explicit about their inputs and outputs using the Kale SDK.
  • Run PyTorch distributed training jobs using Kale pipelines.
  • Kale images with GPU support.
  • Kale-powered VSCode Notebook Image.
  • Upgrade CSI Snapshot CRDs from v1alpha1 to v1beta1.
  • Ship Kale notebooks images with Python 3.8.
  • Support for JubyterLab version 3.

Improvements

  • Improve the MiniKF dashboard.
  • Remove the Contributors tab from the Kubeflow central dashboard.
  • Improve the KFP caching mechanism.
  • Remove critical CVEs from Rok and Kubeflow images.
  • Improve Rok root controller to accept hundreds of new connections more
    efficiently.

Fixed Issues

  • Fix a systemd issue that could make MiniKF fail on boot.
  • Fix a bug in Rok that could lead to snapshots failing after some time, and making pipeline runs go into error.
  • Fix a bug in the Rok CSI driver that could lead to potential failure when trying to create a volume using a VolumeSnapshot as source.
  • Fix an upstream kernel bug that could lead to failed snapshots.
  • Fix race on garbage collection tasks.

You can find more information about bugfixes and improvements here: https://docs.arrikto.com/release-1.5/NEWS.html

MiniKF is the simplest way to get started with Kubeflow and Rok on any platform

Turbocharge your team’s Kubeflow and MLOps skills with a free workshop