mccheah commented on issue #23347: allow building spark gpu docker images URL: https://github.com/apache/spark/pull/23347#issuecomment-453708580 > In comparison, other packages (e.g. instrumentation tools) can be easily added as a new layer on top of the base Spark image. Because we're not publishing image binaries to Dockerhub, the work required to build a custom image with instrumentation tools is the same amount of work required to change the base image: one creates a Dockerfile, or else modifies the provided Dockerfile in-place, and changes the base image. One can see that being done [here](https://github.com/palantir/spark/blob/master/spark-docker-image-generator/src/main/java/org/apache/spark/deploy/kubernetes/docker/gradle/GenerateDockerFileTask.java#L77), for example (as I've linked above). So the question is what makes creating a custom image for GPU support that much more difficult than creating a custom image for generally running on Ubuntu or even running on a Fedora image, etc.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
