LibofRelax commented on issue #26485:
URL: https://github.com/apache/beam/issues/26485#issuecomment-1675640987

   I totally agree that we need better documentation of portable runners in 
general. I've been digging through python and go SDK code for days trying to 
get portable runners to work and OMG the official documentation was bad.
   
   It seems like you misunderstood LOOPBACK environment. Loopback starts a 
server on your host machine (not the flink node) and is solely for debugging 
locally. The beam generated code running on flink will connect to your go code 
on your machine to run UDFs, so it's bounded by local memory.
   
   As for DOCKER environment, the beam code will try to run the 
apache/beam_go_SDK image on flink node so you will need to be able to run 
docker on that machine. To run docker in docker containers, you need to set 
`privileged: true` in your docker-compose config. Although it would be nice to 
run the SDK container as a side car using the EXTERNAL environment option in 
containerized environments such as k8s, it looks like the go SDK has not 
provided worker pool in SDK harness container yet unlike python and java SDKs.
   
   Another important thing to note, you need to mount the job server artifact 
directory to SDK harness containers as they expect to find artifacts in the 
same directory.
   
   I'm not a beam expert though, please take these with a grain of salt.


-- 
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]

Reply via email to