lostluck opened a new issue, #24931:
URL: https://github.com/apache/beam/issues/24931

   ### What happened?
   
   In working on adding process continuation support for #24789 , I determined 
that there's a bug in the Go SDK's implementation of continuation handling.
   
   Essentially, if there's more than a single element in the data source for a 
bundle, then the checkpoint can be lost and overwritten by a subsequent 
element. This prevents checkpointed elements from being re-run in a later 
bundle.
   
   Checked against Java and Python, the correct process is to cache all 
checkpointed elements and return a list of multiple checkpoints to the runner.
   
   The solution is to begin building up the list of delayed elements and their 
checkpoints immediately after initial processing, and not back track later, and 
build up and return a slice of them.
   
   A fix is forthcoming.
   
   ### Issue Priority
   
   Priority: 1 (data loss / total loss of function)
   
   ### Issue Components
   
   - [ ] Component: Python SDK
   - [ ] Component: Java SDK
   - [X] Component: Go SDK
   - [ ] Component: Typescript SDK
   - [ ] Component: IO connector
   - [ ] Component: Beam examples
   - [ ] Component: Beam playground
   - [ ] Component: Beam katas
   - [ ] Component: Website
   - [ ] Component: Spark Runner
   - [ ] Component: Flink Runner
   - [ ] Component: Samza Runner
   - [ ] Component: Twister2 Runner
   - [ ] Component: Hazelcast Jet Runner
   - [ ] Component: Google Cloud Dataflow Runner


-- 
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