Release Notes for DC/OS Data Science Engine version 1.0.1
DC/OS Data Science Engine New Features
- Added automation for enabling/disabling verbose logging for Spark resource offers
- Notebook Docker image is configurable via service options.
- DC/OS Data Science Engine now stores fetched Hadoop configuration files, hdfs-site.xml and core-site.xml, in
$MESOS_SANDBOX/hadoop_conf
folder so users can modify them without root privileges.
DC/OS Data Science Engine Bug Fixes
- Fixed Spark configuration dialog in BeakerX kernel.
- Fixed folder permissions so users can install new packages with
conda
orpip
. - Fixed static resources paths in host mode for components with UI: Spark, History Server, Tensorboard
DC/OS Data Science Engine Breaking Changes
- CUDA 9 images will no longer be provided with a bundle; CUDA 10 is the only version used for GPU images.
jaas_secret
option removed from security configuration.extra_spark_secrets
should be used instead.
DC/OS Data Science Engine Limitations
- DC/OS Data Science Engine does not fully support
root
service user. It is recommended to use the default usernobody
. - DC/OS Data Science Engine does not support the installation of Python packages on CoreOS. It is recommended to use CentOS.
- DC/OS Data Science Engine does not support AWS Classic Load Balancer, because it does not support WebSockets. It is recommended to configure load balancers and proxies to allow WebSockets.