Identity Resolution¶
UNALLOCATED identity¶
If a billing line item cannot be resolved to any identity, cost is attributed to the
system UNALLOCATED identity (type=system). This is created automatically by the
orchestrator on first run.
Confluent Cloud¶
Identity types: service_account, user, identity_pool
Resolved via CCloud API — principals linked to Kafka clusters via ACLs and metrics. Identity resolution uses metrics data (bytes in/out per principal) to weight splits.
Self-managed Kafka¶
Identity types: principal, team
Three discovery modes:
identity_source.source |
How identities are found |
|---|---|
prometheus |
Extracted from Prometheus metric label principal |
static |
Loaded from static_identities list in config |
both |
Union of Prometheus-discovered + static |
principal_to_team maps raw principal IDs to team names.
Generic metrics¶
Identity types: configurable via identity_source.label
Same three modes as self-managed Kafka. Label name is configurable — not fixed to
principal.
Allocation strategies¶
| Strategy | When to use |
|---|---|
even_split |
Infrastructure costs that benefit all users equally |
usage_ratio |
Usage-driven costs (network, compute by workload) |
Even split: cost ÷ identity count.
Usage ratio: each identity gets their_usage / total_usage × cost.