Closed wjglerum closed 5 years ago
Hi @wjglerum, just to clarify prod-dist
step is used to checkout Apache Spark sources and build a dist tarball for the future use with Dockerfile. If you're looking into creating a stub and deploy it to your DC/OS cluster, consider using publish_local_spark.sh script instead. Master branch of this repo depends on Spark 2.4 distribution.
Are you using a custom SPARK_DIR
for the build? Is it an absolute path? Also, which version of make
are you using?
If SPARK_DIR
is not provided then mesosphere/spark
master branch will be used which is a bit off of the upstream. You can try to checkout custom-branch-2.4.x which has all the latest changes and tested with all supported DC/OS versions.
So the whole process should look that:
mkdir dcos-spark
cd dcos-spark
export DIR_PATH=$(pwd)
git clone git@github.com:mesosphere/spark.git
cd spark
git checkout custom-branch-2.4.x
cd ..
git clone git@github.com:mesosphere/spark-build.git
cd spark-build
SPARK_DIR=${DIR_PATH}/spark make prod-dist
Hi, thank you for your reply. I will try to test it again. For now Spark 2.4.0 is available already in the catalog so we are fine ;)
I tried building this project locally to give spark 2.4.0 a try. I followed the README, however the Makefile doesn't work for me. Did I miss anything?
make docker-build
works and a container is createdmake prod-dist
fails with: