retention.bytes Required Alongside retention.ms

Every non-compacted topic must declare a retention.bytes cap.

Rationale

Time-only retention is unbounded in disk. A 10x traffic spike fills the broker before the time window expires and the cluster goes read-only. Setting both makes the topic self-defending: whichever limit hits first triggers cleanup. Matches the `partitions x retention.bytes` capacity model platform teams use to plan disk.

Pattern

cleanup.policy contains "delete" → retention.bytes set and > 0

Examples

cleanup.policy=delete + retention.ms=604800000 + retention.bytes=10737418240

cleanup.policy=compact (rule skipped)

cleanup.policy=delete + retention.bytes unset → unbounded disk growth

Parameters

Name	Default	Description
`min_bytes`	`1`	Minimum acceptable retention.bytes value.

Governs

This policy relates to the following Kafka configuration keys:

Implementation

Drop this YAML into Conduktor Console as a ResourcePolicy, then link it from an ApplicationInstance, Application, or KafkaCluster.

Conduktor ResourcePolicy

# Conduktor self-service ResourcePolicy
---
apiVersion: self-serve/v1
kind: ResourcePolicy
metadata:
  name: retention-bytes-required
spec:
  targetKind: Topic
  description: delete-policy topics must declare retention.bytes > 0
  rules:
    - condition: '!("cleanup.policy" in spec.configs) || !spec.configs["cleanup.policy"].contains("delete") || ("retention.bytes" in spec.configs && int(string(spec.configs["retention.bytes"])) > 0)'
      errorMessage: "retention.bytes must be set (>0) on delete-policy topics — time-only retention is unbounded in disk"

Related policies

Try Conduktor Console

Enforce policies like this across your team — central audit history, pre-commit guardrails, ApplicationInstance bindings. 5-min Docker install.

Get Started →