Skip to main content
Chargeback is a cost allocation feature that helps you track, visualize and attribute Kafka infrastructure costs to clusters, teams or applications. By measuring resource use over time, Chargeback gives you visibility into Kafka spending, supports cost accountability and helps you make informed decisions about resource optimization. Conduktor measures five cost axes — storage, partitions, ingress, egress and Kafka Connect — and presents costs through three views: by cluster, by application or by label.
From Console v1.45.0, Chargeback unifies the previous with Gateway and without Gateway experiences into a single page. The previous V1 and V2 endpoints are deprecated but still respond to existing API clients.
From our blog: Chargeback: attribute and map Kafka costs to your business Practical strategies to tag, trace, and report Kafka spending by team and application.

​
Cost axes

Each cluster has its own configurable unit costs across five axes. Set the costs that match your provider pricing or your internal infrastructure model.
Cost axisUnitWhat it measures
Storage$/GB/hourBytes retained on disk per topic
Partitions$/partition/hourNumber of partitions per topic
Ingress$/GBBytes produced into Kafka
Egress$/GBBytes consumed from Kafka
Kafka Connect$/task/hourHours that Kafka Connect connector tasks run
A cluster only needs the axes that match its pricing model. For an on-premises cluster, you might only set storage and partitions; for a Confluent Cloud cluster, you can set all five. Kafka Connect costs use a default per-task-hour rate for the cluster. You can also set per-connector-class overrides — for example, a higher rate for an S3 sink connector than for a JDBC source connector. Costs are calculated at query time, so updating a unit cost takes effect immediately across historical data.

​
What you can measure per cluster type

Chargeback collects metrics from four sources: Console’s metadata indexer for storage and partition counts, the Confluent Cloud Metrics API for Confluent Cloud throughput, the observability Interceptor for Gateway throughput, and Console’s Kafka Connect integration for connector task hours.
Cluster typeStoragePartitionsIngress / egress per topicIngress / egress per service accountKafka Connect
Vanilla Kafka✓✓––✓
Confluent Cloud✓✓✓✓ (no topic breakdown)✓
Gateway✓✓✓✓✓
Kafka Connect costs apply to any cluster type, but only when the cluster has one or more Kafka Connect clusters configured in Console. Clusters without Kafka Connect don’t show Connect costs. Storage and partition metrics are bucketed by UTC day. Confluent Cloud throughput is collected hourly. Gateway throughput is streamed continuously through the observability Interceptor and flushed at a configurable interval.

​
Views

Chargeback presents data through three top-level views, each with the same cost columns and the ability to drill down into topics, service accounts and connectors.
  • Cluster view: costs grouped by Kafka cluster
  • Application view: costs grouped by , with drill-down into application instances
  • Label view: costs grouped by the values of a label key (for example, env=production or team=payments)
Label-based grouping requires the label key to first be selected as a Chargeback Label by an administrator.

​
Chargeback Labels

Not every label attached to a topic or is meaningful for cost attribution. Administrators select a subset of label keys to expose in Chargeback. Once a label key is enabled, you can group costs by its values. Topics or service accounts without that label are excluded from the grouping. Chargeback Labels are global — the same set of label keys applies across all clusters.

​
Permissions

Two platform-level permissions control access:
  • Can view Chargeback (PlatformChargebackView): view cost metrics across all views and drill-downs
  • Can manage Chargeback (PlatformChargebackManage): set per-cluster unit costs and select Chargeback Labels
By default, platform admins have both permissions. Console shows cost attribution for all clusters regardless of the cluster-level permissions assigned to the user.