Published by Theofilos Kakantousis on

Introducing Hopsworks 0.7.0

Hopsworks 0.7.0 brings the latest features, improvements and bug fixes. It comes a short while after version 0.6.0 which brought new features such as Model Serving and Experiments. This version paves the way for the next batch of new features and improvements coming in 0.8.0 such as the Feature Store and a revamped REST API.

As part of the Feature Store implementation, Hopsworks 0.7.0 adds support for SparkOnHive, i.e. using SparkSQL to read and write data stored in Apache Hive. Users can now run such queries from within their Jupyter notebooks as per this example. Some more documentation is available here.

You can get started with Hopsworks by visiting our getting started guide.

Release notes – Hopsworks – Version 0.7.0

Bug

  • [HOPSWORKS-713] – Tame setup-chef
  • [HOPSWORKS-801] – Jupyter application view/kill buttons not visible in low resolutions
  • [HOPSWORKS-803] – Add Elastic dependency to Berksfile in hops-hadoop-chef
  • [HOPSWORKS-805] – Add rm dependency to nm in hops-hadoop-chef karamelfile
  • [HOPSWORKS-807] – fix yarn.nodemanager.aux-services in mapred-site.xml
  • [HOPSWORKS-808] – fix hops cluster definition in karamel-chef
  • [HOPSWORKS-812] – TensorBoard poller running too often
  • [HOPSWORKS-815] – Mail sent from dataset request addresses wrong user
  • [HOPSWORKS-820] – Only one instance of JAXBContext should be created per JobConfiguration.
  • [HOPSWORKS-823] – Jupyter cleanup for Project deletion never worked
  • [HOPSWORKS-824] – [conda-chef] Template conda repo only if provided for hops-system environment
  • [HOPSWORKS-825] – Review projectmember removal and jobs
  • [HOPSWORKS-830] – conda hops-system env fails the upgrade from 0.6.* to 0.7.0
  • [HOPSWORKS-833] – NPE when getting the state of CertificateMaterializer
  • [HOPSWORKS-843] – TensorBoard cleans up certificates twice
  • [HOPSWORKS-847] – Add Sparkmagic support for utf-8 chars in configuration file

Sub-task

Improvement

  • [HOPSWORKS-679] – [kagent-chef] Use Anaconda environment for kagent
  • [HOPSWORKS-728] – Support Random Search in hops-util-py
  • [HOPSWORKS-750] – Jupyter server should use the project conda env
  • [HOPSWORKS-767] – Support for installing multiple library versions with pip
  • [HOPSWORKS-780] – [ePipe] InodeIds are long starting from HopsFS 2.8.2.6
  • [HOPSWORKS-793] – TensorBoard in Experiments service too long lived
  • [HOPSWORKS-800] – Commands executor service
  • [HOPSWORKS-810] – Avoid launching a subprocess to read Hadoop classpath
  • [HOPSWORKS-811] – Lazy Conda environment creation
  • [HOPSWORKS-813] – Speed up TensorBoard in Experiments service
  • [HOPSWORKS-831] – [ePipe] Use Uint64 as the processing index instead of long

Task

Categories: