Create a Kafka Topic: Partitions, Replication, and Retention

From the Topics page, click Create Topic. Start with a preset, then adjust the settings for the workload and connected cluster.

Choose a Kafka topic preset

New Topic opens with four presets. Each one is a starting point you can adjust, not a mode you are locked into.

Creating a Kafka topic from a preset with partitions, replication, retention, cleanup policy, and advanced configuration

Preset	Configuration	When it fits
Standard	3 partitions, 3 replicas, Delete, 7-day retention	General use on a multi-broker cluster
Development	1 partition, 1 replica, Delete, 30-day retention	Local or temporary environments
Compacted	3 partitions, 3 replicas, Compact	Keyed state and changelog data
High Throughput	12 partitions, 3 replicas, Delete, 12-hour retention, 1 GB per partition	Higher parallelism with bounded storage

Kafka topic naming conventions

Use a unique name of up to 249 characters. Kafma accepts letters, numbers, dots, underscores, and hyphens; spaces are not allowed.

Avoid mixing dots and underscores across topic names. Kafka's metric names replace dots with underscores, so orders.created and orders_created can collide in monitoring even though they are different topics.

Choose partitions and replication factor

Partitions

The partition count limits how many members of one consumer group can read the topic in parallel. More partitions can increase throughput, but also add storage, replication, and rebalance overhead.

Kafka lets you add partitions later, but not remove them. Adding partitions can also change how keyed records are routed. See Adding Kafka partitions before increasing the count.

Replication factor

The replication factor cannot exceed the cluster's broker count. A factor of 1 stores no redundant copy, so another replica cannot take over when that broker is unavailable. Higher values improve resilience but use more storage and replication traffic.

Presets with three replicas require at least three brokers. Reduce the value when creating the topic on a smaller cluster.

Retention and cleanup policy

For policies that include Delete, retention time controls when old records become eligible for deletion. Cleanup policy decides what "cleaning up" means:

Delete removes old segments according to the topic's time and size limits.
Compact retains the latest value for each key as Kafka cleans segments.
Mixed applies both deletion and compaction.

Create a compacted Kafka topic

Choose the Compacted preset or set the cleanup policy to Compact. Compacted topics work by key, so producers should include one. A record with a null value is a tombstone that marks its key for removal.

Compaction is not immediate. It works on closed segments and is governed by the broker's cleaner settings, so a duplicate key can remain visible for a while after the newer record is written.

Advanced Kafka topic configuration

Advanced Configuration exposes three common topic overrides:

min.insync.replicas sets how many in-sync replicas must acknowledge a write when the producer uses acks=all.
retention.bytes limits retained data per partition. The total across the topic depends on its partition count.
max.message.bytes sets the largest record batch the topic accepts and appears as the message limit in the topic details drawer.

Use Add Custom Property for another Kafka topic configuration.

Edit Kafka topic configuration

Open a topic's action menu and select Edit Configuration to change retention, cleanup policy, and topic-level overrides. Clearing an override resets it to the broker default.

Partitions and replication factor are shown but cannot be changed in this form. Use Add Partitions to increase the partition count. Kafma does not currently provide replica reassignment.