Skip to main content

Interactions between Dask and Spark

Project description

Launch Dask from Spark and Spark from Dask. This project is not mature.

Examples

pip install dask-spark

Create Spark cluster from a Dask cluster

>>> from dask.distributed import Client
>>> client = Client('scheduler-address:8786')
>>> client
<Client: scheduler='tcp://scheduler-address:8786' processes=8 cores=64>

>>> from dask_spark import dask_to_spark
>>> sc = dask_to_spark(client)
>>> sc
<pyspark.context.SparkContext at 0x7f62fa4bb550>

Create Dask cluster from a Spark cluster

>>> import pyspark
>>> sc = pyspark.SparkContext('local[4]')
<pyspark.context.SparkContext at 0x7f8b908b0128>

>>> from dask_spark import spark_to_dask
>>> client = spark_to_dask(sc)
>>> client
<Client: scheduler="'tcp://127.0.0.1:8786'">

Requirements and How this Works

This depends on a relatively recent version of Dask.distributed.

For starting Spark from Dask this assumes that you have Spark installed and that the start-master.sh and start-slave.sh Spark scripts are available on the PATH of the workers. This starts a long-running Spark master process on the Dask Scheduler and starts long running Spark slaves on Dask workers. There will only be one slave per worker. We set the number of cores and the amount of memory to match the Dask workers and available memory.

When starting Dask from Spark this will block the Spark cluster. We start a scheduler on the local machine and then run a long-running function that starts up a Dask worker using RDD.mapPartitions.

TODO

  • [ ] This almost certainly fails in non-trivial situations
  • [ ] Enable user specification of Java flags for memory and core use
  • [ ] Support multiple spark clusters per Dask cluster

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
dask_spark-0.0.2-py2.py3-none-any.whl (5.3 kB) Copy SHA256 hash SHA256 Wheel py2.py3
dask-spark-0.0.2.tar.gz (3.6 kB) Copy SHA256 hash SHA256 Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page