apache / pinot

Apache Pinot - A realtime distributed OLAP datastore
https://pinot.apache.org/
Apache License 2.0
5.53k stars 1.3k forks source link

Handle change in default groupId behavior in kafka consumer #9183

Open navina opened 2 years ago

navina commented 2 years ago

Starting kafka 2.2.0 onwards, when group.id is not specified in the kafka consumer config, it is set as null, instead of an empty string. Due to this change, even though auto-commit is enabled, it doesn't show up in the consumer offsets topics. This breaks existing kafka monitoring tools.

One proposed solution is for Pinot to set a default group id when it was not explicitly configured by user (say, table name as the default group id).

navina commented 2 years ago

@npawar please tag as beginner-task

tanmesh commented 2 years ago

I am planning to pick this task.

tanmesh commented 2 years ago

Hey @npawar @navina 👋 , can you please assign someone to review this PR 🙏

pjpringle commented 2 years ago

I didn't think pinot should be using Kafka consumer group offset management nor auto.commit as pinot manages it own consumer partition assignment and offsets. Using a group id results in a subset of partitions getting assigned to each instance and stops the replica count from working.