I’ve been reconciling Confluent Cloud invoices line by line, mapping every line_type in the billing CSV to what it actually meters. I’d like to check my understanding before relying on it.
| line_type | I think it bills | Unit |
|---|---|---|
KAFKA_NUM_CKUS / KAFKA_NUM_ECKUS |
Whole-cluster capacity | CKU-hour |
KAFKA_STORAGE |
Retained bytes; replication already included | GB-hour |
KAFKA_NETWORK_READ / _WRITE |
Network egress / ingress | GB |
KAFKA_PARTITION |
Partitions beyond the free allowance (Basic/Standard only?) | partition-hour |
CONNECT_NUM_TASKS |
Connector tasks, whether moving data or not | task-hour |
CONNECT_THROUGHPUT |
Bytes processed by the connector | GB |
FLINK_NUM_CFU |
Flink compute | CFU-hour |
GOVERNANCE_BASE |
Schema Registry / Stream Governance | hour |
A few things I’m unsure about:
- Storage and replication: Is
KAFKA_STORAGEmeasured as logical retained bytes, with RF3 already reflected in the price? My assumption is that replication-factor changes cannot reduce this cost in Confluent Cloud because RF is fixed at 3. - Dedicated partitions: I don’t see
KAFKA_PARTITIONfor Dedicated clusters. Does that mean partitions are covered by CKU capacity rather than billed separately? - Connector tasks: Does
CONNECT_NUM_TASKSbill for provisioned connector tasks regardless of whether the connector is actually moving bytes? In other words, would a paused—but not deleted—connector still incur task-hour charges? - Negative lines / credits: Do credits or adjustments ever appear as negative lines in the CSV, or are they handled elsewhere?
If you’ve reconciled an invoice against these fields, I’d appreciate any confirmation or corrections.