pulsar-spark
pulsar-spark copied to clipboard
[BUG] unnecessary subscription created when processing batches
Describe the bug
Why do we need to create subscription when reader API is used and Spark manages offsets on its own. I can see new subscriptions got created and deleted by PulsarMetadataReader before and after committing a batch but dont see it useful. Is that legacy code and should be removed ?
Expected behavior
- No subscription creation needed when reader API is used
Additional context Spark version 2.4.5 and Pulsar version 2.5.0