aaltay commented on pull request #15105: URL: https://github.com/apache/beam/pull/15105#issuecomment-912884587
I do not believe we should make this change. And this is not related to Google's private packages. The reasons are: - Staging packages creates a consistent set of artifacts throughout the lifetime of a pipeline. (Consider a long running streaming job, with autoscaling etc, different works might be started with days of gap). Downloading from a location other than staging location could result in different dependencies in different workers (e.g. a package getting an update in pypi). - If users want to do this they can do it by using a custom containers. Beam's custom container protocol (https://s.apache.org/beam-fn-api-container-contract) and allows for a completely custom container with support for changes like this (https://beam.apache.org/documentation/runtime/environments/) - It is also possible to achieve the results here even without using custom containers. (This is option is good for legacy Dataflow pipelines as well) by using a custom commands setup.py file (https://beam.apache.org/documentation/sdks/python-pipeline-dependencies/#nonpython). This file can execute any code at worker startup time including downloading and installation of arbitrary dependencies. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
