Question
Does an empty Kafka or Event Hubs partition assignment mean that the Cribl Source is unhealthy or losing data?
Environment
- Cribl Stream
- Kafka or Kafka-compatible Event Hubs Source
- Consumer group with more consumers than partitions for a topic
Answer
No. A consumer can legitimately receive an empty partition assignment when the topic has fewer partitions than the total number of consumers.
An empty assignment for one consumer does not by itself prove data loss or indicate that the Source is unhealthy. Other consumers may hold the available partitions and continue processing data.
Evaluate the source as a whole by checking:
- Whether any consumers received non-empty assignments.
- Whether Events In is non-zero.
- Whether consumer lag changes as expected.
- Whether the group is repeatedly rebalancing.
- Whether heartbeat, metadata, or fetch errors are present.
- Whether all consumers are idle or only a subset has no partitions.
Size the worker group for aggregate workload, not for a one-to-one partition ratio for a single topic.
Additional Information
If all consumers are idle or the source has zero Events In, investigate broker connectivity, repeated rebalances, worker initialization time, disk I/O, and source configuration separately.