[ 
https://issues.apache.org/jira/browse/BEAM-11152?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17486818#comment-17486818
 ] 

Anand Inguva edited comment on BEAM-11152 at 2/4/22, 5:18 AM:
--------------------------------------------------------------

You can find relevant details in how to use custom base image that works with 
the dataflow workers here.

[https://cloud.google.com/dataflow/docs/guides/using-custom-containers#using_a_custom_base_image_or_multi-stage_builds]

Note: The entry point of the image should be */opt/apache/beam/boot* and needs 
to follow the instructions on the above link. The entry point MUST be set. If 
not,  spinning up the dataflow workers with your image as base image would 
fail. 

These instructions will be added to the Beam Python page soon. 

 


was (Author: anand inguva):
You can find relevant details in how to use custom base image that works with 
the dataflow workers here.

[https://cloud.google.com/dataflow/docs/guides/using-custom-containers#using_a_custom_base_image_or_multi-stage_builds]

Note: The entry point of the image should be */opt/apache/beam/boot* and needs 
to follow the instructions on the above link. The entry point MUST be set. If 
not,  spinning up the dataflow workers with your image as base image would 
fail. 

 

> publish guide or scripts to build Beam SDK harness container from an existing 
> container image
> ---------------------------------------------------------------------------------------------
>
>                 Key: BEAM-11152
>                 URL: https://issues.apache.org/jira/browse/BEAM-11152
>             Project: Beam
>          Issue Type: Improvement
>          Components: sdk-py-harness
>            Reporter: Zhitao Li
>            Priority: P3
>
> The current image building doc in 
> [https://beam.apache.org/documentation/runtime/environments/#customizing-container-images]
>  only provides paths of modifying existing Dockerfile or extending the image 
> path.
>  
> It might be the case that certain large enterprise users have their own 
> "blessed" base images which are widely extended within their company, and 
> teams might want to adapt these images to fit into Dataflow workers. 
> Publishing guidelines on this path might be helpful.
>  
> Even better, if a python equipped container images installs `apache-beam` (or 
> an extra dependency) python package, it should be possible to be used as a 
> container image.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to