Transaction coordinator

As explained in Apache Kafka in Action

A transaction coordinator is the Kafka broker component that coordinates a transaction and writes markers recording its commit or abort outcome.

A transaction coordinator is the Kafka broker component that manages a transaction across partitions. Producers register target partitions, write tentative messages, and then report completion. The coordinator records the outcome with commit markers or abort markers, preserving all-or-nothing behavior even when brokers or producers crash. section 5.2.3

Flight and hotel reservation

A flight-and-hotel booking illustrates the flow: reserve the hotel with cancellation available, book the flight, then confirm the hotel only after the flight succeeds. Kafka similarly holds messages as tentative until the coordinator commits them; an abort marker causes consumers to skip them. section 5.2.3

A read_committed consumer does not need to understand transaction state. The broker holds messages after an unfinished transaction to preserve ordering, releases them after a commit marker, and skips aborted messages while continuing with unrelated records when permitted. section 5.2.3

Figure 5.7 Our consumer stops receiving new messages as soon as a transaction remains unfinished because the broker only sends messages up to the start of the unfinished transaction and holds back all subsequent messages, even those unrelated to the transaction. The ordering guarantee still holds. When a transaction-commit is received, all messages are released and sent to our consumer. If a transaction is aborted, our consumer doesn’t receive messages belonging to the aborted transaction, but messages not belonging to the transaction are still received normally.

Follow the message flow through unfinished, committed, and aborted transactions to see when the broker releases, delays, or skips records. section 5.2.3

How Apache Kafka in Action covers it

How deep it goes

Explains partition registration, tentative writes, outcome markers, crash guarantees, and broker-side read_committed filtering.

Examples

Uses a flight-and-hotel reservation analogy and a message-flow figure covering unfinished, committed, and aborted transactions.

What it assumes

Assumes familiarity with Kafka partitions and consumers while explaining transaction visibility through concrete booking and message-flow examples.

What you can do after

Evaluate transaction reliability and understand how Kafka consumers receive only committed records.

How it connects

Connects transactions with idempotent producers, two-phase commit, partitions, and read_committed consumers.

liveBookmanning.com

Read Apache Kafka in Action on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Think Distributed SystemsA component that receives requests, assigns them an order, and distributes the ordered requests to other components.

Compare how 2 books explain transaction coordinator →

Related concepts

• Kafka brokerkind of · A Kafka broker is a server process in an Apache Kafka cluster that stores records and handles producer, consumer, and other client requests.
• distributed transactionused by · A distributed transaction is a unit of work whose operations span multiple resources or systems and whose outcome is coordinated across them.
• transactional producerused by · A transactional producer is a Kafka producer mode that publishes records to multiple topics as one atomic transaction, committing all records together or none of them.
• race conditionaddresses · A race condition is a concurrency defect in which program behavior or results depend on the relative timing or ordering of interacting operations.

This explainer was generated by Manning's knowledge graph from the text of Apache Kafka in Action. The section links point to the parts of the book it draws on.