Dear Apache Beam Team,
I am a student at the University of Innsbruck, currently using Apache
Beam for my research.
I am using Flink v1.19 and the Flink runner
(https://beam.apache.org/documentation/runners/flink/) to execute
pipelines. I managed to run the pipeline with several options:
- parallelism
- operatorChaining
- attachedMode
- ...
All of these options were applied correctly, and I could immediately
see their effects.
However, the "maxBundleTimeMills" and "maxBundleSize" settings do not
seem to work properly. I tried setting them, but I observed no effect
when gathering metrics. I created an artificial pipeline where I set
maxBundleTimeMills=100000 (100 seconds) and maxBundleSize=10000
elements. I configured a Kafka topic that gets 1 element/second. I
expected to receive a whole bundle of messages after 100 seconds.
However, while measuring, I noticed that I actually receive 1 message
per second continuously, indicating that these two settings are not
being applied correctly. Am I right or missing something?
Could you provide me with assistance in solving this issue?
I also tried different Flink versions (e.g., 1.16.3).
It would be wonderful if I could join the Slack community. Could you
provide me with an invitation, or is it restricted to @apache.org users?
Thank you very much for your help, and I look forward to your reply.
Best regards,
Stefan Pedratscher