Skip to main content

Partition segments

Apache Kafka® divides topics partition data into segment files (with .log suffix) stored on the file system. Each segment file is named using the offset of the first message (a.k.a. base offset) contained. For example, the segment file 04.log contains the message with offset 4 as first entry.

The last segment in the partition is called the active segment and it is the only segment to which new messages are appended to.

In the example above the topic partition is divided into multiple segments, with 06.log being the active one:


110014 Privet Dr
21002221B Baker Street
31003Milkman Road


4100221 Jump St
51001Paper St

06.log (active segment)

61001Paper Road 21

When the segment file reaches a certain size or age, Apache Kafka will create a new segment file. This can be controlled by the following settings:

  • segment.bytes : creates a new segment when current segment becomes greater than this size. This setting can be set during topic creation and defaults to 1GB.
  • : forces the segment to roll over and create a new one when the segment becomes older than this value.

Aiven for Apache Kafka® service is fully manageable, the partitions are rebalanced automatically when the cluster is scaled up or down.