Scale Capacity

Caracal publishes signals and Helm controls, but no universal throughput, latency, or availability figure. Establish a baseline with your workload. For measured reference results from the v0.2.0 validation under stated test conditions, see Performance and Scalability.

Decision Criteria

Signal	Investigate first
Database pool pressure	Slow queries, transactions, connections, I/O
Audit lag, DLQ, replay age	Audit write path, Redis, Postgres, replay storage
Gateway STS circuit	STS readiness and exchange latency
Revocation lag	Redis latency, pending entries, snapshot refresh
Readiness flapping	CPU, memory, dependencies, probes

Prerequisites

Collect authenticated metrics, latency, dependency metrics, logs, and repeatable load. Know storage connection/memory ceilings.

Safe Procedure

1. Record baseline and first saturated resource.
2. Correct dependencies or slow queries before replicas.
3. Change one limit or replica count at a time.
4. Keep aggregate pools within database capacity.
5. Repeat load and compare readiness, errors, lag, DLQ, replay, and revocation.
6. Set environment thresholds from measured bounds.

Chart replica/HPA defaults are not capacity recommendations.

Rollback

Restore the previous value if pressure or errors increase and retain both measurements.

Next Step

Build gates in Monitor Health and Metrics.