KIP-932 Share Group Visualizer

Topology

partitions 1

1369

consumers 1

136912152125

ingress rate (events/sec) 50

50100150300500

show producer batches

enabled

Groups records by the producer batch that created them (batch.size = 10). Vertical lines mark where one producer batch ends and the next begins. Useful for observing how share.acquire.mode=batch_optimized aligns — or doesn't align — consumer acquisitions to producer batch boundaries.

Consumer Config

share.acknowledgement.mode

implicit — records are automatically acknowledged when poll() is called. No explicit ack needed.

explicit — the application must call acknowledge(record, AcknowledgeType) for each record. Enables ACCEPT, RELEASE, REJECT, and RENEW outcomes.

max.poll.records 3

1510

The maximum number of records returned in a single poll() call. Higher values mean larger batches per consumer per tick, increasing throughput but reducing fairness across consumers.

share.acquire.mode

batch_optimized — the broker may return fewer than max.poll.records to optimise batching efficiency (KIP-1206). With Show producer batches enabled, observe how acquisitions can align to the batch boundary markers.

record_limit — strictly enforces max.poll.records; acquisitions may cross producer batch boundaries.

processing time (sec/batch) 1s

1s2s3s5s10s15s25s

Delivery Config

share.delivery.count.limit 5

2510

share.renew.acknowledge.enable

enabled

Space — pause / resume

Apache Kafka 4.0 introduced Share Groups (KIP-932), enabling true queue semantics alongside Kafka's existing stream processing model. Unlike consumer groups, multiple consumers in a share group can compete for records from the same partition — making it straightforward to scale processing without repartitioning.

This visualizer focuses on the Share Partition state machine: how individual records move through AVAILABLE → ACQUIRED → ACKNOWLEDGED → ARCHIVED, how the Share Partition Start Offset (SPSO) and End Offset (SPEO) advance, and how delivery count limits, acknowledgement modes, and consumer scaling interact.

⏸ PAUSED — hover over records to inspect

How to Use

Use the Topology controls to set partition count and the number of competing consumers.
Switch share.acknowledgement.mode to explicit to see RELEASE, REJECT, and RENEW acknowledgement types in action.
Lower share.delivery.count.limit to 2 or 3 to watch records hit the ARCHIVED state quickly after repeated redelivery.
Hit ⏸ Pause (or press Space), then hover over any record cell to inspect its partition, offset, state, delivery count, and lock status.
Click ⚡ Crash Consumer to simulate a session timeout — acquired records return to AVAILABLE without acknowledgement.
Click the i icon next to any setting for a description from the Kafka documentation.

Learn More

▶ Queues for Kafka Explained (KIP-932)
Sandon Jacobs · Confluent Developer
✦ How Apache Kafka Flexed to Support Queues
Sandon Jacobs · InfoWorld
⬡ Apache Kafka 4.3 Documentation
Share Groups · kafka.apache.org
⬡ KIP-932: Queues for Kafka
Apache Kafka Wiki

Share Group Visualizer KIP-932 · Apache Kafka 4.3

Share Group Visualizer