By the numbers

A quantitative snapshot of the Apache Kafka codebase. Counts come from the working tree at the wiki's commitHash (see .wiki-meta.json) and from git log against the default branch (trunk). Languages and modules listed here mirror what the build system tracks.

Data collected on 2026-04-30.

Size

xychart-beta horizontal
    title "Lines of code by language"
    x-axis [Java, Scala, Python, Other]
    y-axis "kLOC" 0 --> 1500
    bar [1444, 150, 50, 30]

Language	Files (approx)	Lines (approx)
Java	5,878	1,443,763
Scala	296	149,699
Python	177	system tests + helpers in tests/, committer-tools/, vagrant/
Shell	~90	launcher scripts under bin/, tests/, docker/
JSON schema	~250 RPC messages	under clients/src/main/resources/common/message/ and per-coordinator schema dirs

Java is the dominant language; Scala is concentrated in core/ (the broker request-handling layer) and in streams/streams-scala/.

Modules

The Gradle build defines roughly 50 modules in settings.gradle. The most active are:

Module	Purpose
clients	Producer, Consumer, Admin, shared common/* types
core	Scala broker — KafkaApis, ReplicaManager, request pipeline
streams	Kafka Streams library
connect/runtime	Connect framework
raft	KRaft consensus implementation
metadata	Metadata image, controller, MetadataLoader
storage, storage-api	Log storage, segments, indexes, tiered storage
group-coordinator	Consumer group coordinator (classic + KIP-848)
share-coordinator	Share group coordinator (KIP-932)
transaction-coordinator	Transaction coordinator
coordinator-common	Shared replicated-state-machine runtime for all coordinators
tools	Java implementations of kafka-*.sh CLIs
server, server-common	Shared broker infrastructure used outside core
jmh-benchmarks	JMH microbenchmarks
trogdor	Distributed test/fault-injection harness
streams:upgrade-system-tests-*	One module per supported Streams upgrade path (0.11 through 4.x)

Activity

Window	Commits
Total (since 2011-08)	17,420
Last 90 days	518
2024 (calendar year)	2,448
2025 (calendar year)	2,185

xychart-beta horizontal
    title "Commits per year (default branch)"
    x-axis [2011, 2013, 2015, 2017, 2019, 2021, 2023, 2025]
    y-axis "commits" 0 --> 2500
    bar [116, 350, 762, 1507, 1219, 1280, 1474, 2185]

The trend roughly tracks Kafka's adoption: a bump in 2017 (Streams + Connect maturing), elevated activity through the KRaft transition (2020-2024), and very high activity in 2024-2025 driven by the ZooKeeper removal, KIP-848 (new consumer protocol), KIP-932 (share groups), and KIP-405 (tiered storage) work.

Churn hotspots (last 90 days)

The files touched most often in the last 90 days are concentrated in the new consumer-group protocol and the broker request handling layer:

File	Touches
group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupMetadataManager.java	26
group-coordinator/src/test/java/org/apache/kafka/coordinator/group/GroupMetadataManagerTest.java	22
core/src/test/scala/unit/kafka/server/KafkaApisTest.scala	22
gradle/dependencies.gradle	18
build.gradle	17
core/src/main/scala/kafka/server/KafkaApis.scala	15
streams/test-utils/src/main/java/org/apache/kafka/streams/TopologyTestDriver.java	13
group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupConfig.java	13

That signal lines up with what's currently shipping: the consumer-group state machine, the Scala broker's RPC dispatcher, and dependency upgrades.

Contributors

The wider Kafka community is large; the snapshot below is just the most active committers (by number of commits to the default branch since 2024-01-01) — useful for triage rather than for ownership. There is no CODEOWNERS file, so subsystem ownership is best discovered via git log <path> (see maintainers.md — not generated; ownership is tracked via the Apache committer roster instead).

Contributor	Commits since 2024-01-01
PoAn Yang	239
Ken Huang	238
Matthias J. Sax	218
David Jacot	201
TengYao Chi	175
Andrew Schofield	165
Mickael Maison	160
David Arthur	132
Lucas Brutschy	130
Apoorv Mittal	124

Bot-attributed commits

A git log --grep for the standard bot signatures (factory-droid[bot], dependabot[bot], github-actions[bot], copilot[bot]) finds 1 matching commit out of 17,420 total. Apache projects historically merge bot work under human committer identities, so this is a hard lower bound on AI- or bot-assisted activity rather than a measurement of it.

Complexity hotspots

A few files are notably large and worth knowing about — they tend to be the ones a contributor encounters first when changing broker behavior, the consumer, the producer, or Streams.

File	Size (lines)
core/src/main/scala/kafka/server/KafkaApis.scala	~5,200
core/src/main/scala/kafka/server/ReplicaManager.scala	~3,600
clients/src/main/java/org/apache/kafka/clients/consumer/KafkaConsumer.java	~3,000
streams/src/main/java/org/apache/kafka/streams/KafkaStreams.java	~2,700
clients/src/main/java/org/apache/kafka/clients/producer/KafkaProducer.java	~2,500
streams/src/main/java/org/apache/kafka/streams/StreamsConfig.java	~3,100
connect/runtime/src/main/java/org/apache/kafka/connect/runtime/Worker.java	~3,400
raft/src/main/java/org/apache/kafka/raft/KafkaRaftClient.java	~4,400

These files are also the most expensive to review and the most touched. Several of them are split into smaller helpers in their internals/ package — for example, the producer's Sender, RecordAccumulator, and TransactionManager carry most of the actual logic that KafkaProducer.java orchestrates.