> ## Documentation Index
> Fetch the complete documentation index at: https://docs.doblier.io/llms.txt
> Use this file to discover all available pages before exploring further.

# On-Call Runbook

> Every recovery time in an ORCA runbook is measured in your environment — never assumed.

An ORCA runbook is **generated from evidence**: the mandatory [chaos suite](/orca/operations/chaos-testing) injects seven faults against production-sized state in your environment, and the documented recovery times become your on-call runbook. These are the reference targets a deployment must meet before go-live:

| # | Scenario                  | Measured recovery target                |
| - | ------------------------- | --------------------------------------- |
| 1 | Redis master lost         | Failover \< 10s, zero event loss        |
| 2 | Neo4j replica lost        | Enrichment continues, latency +\< 100ms |
| 3 | ServiceNow webhook outage | Fallback poll engages \< 60s            |
| 4 | Flink task-manager crash  | Checkpoint recovery \< 60s              |
| 5 | Schema Registry down      | Local cache serves 30 min               |
| 6 | ClickHouse node lost      | HA failover, 0 records lost             |
| 7 | Kafka broker lost         | Rebalance \< 30s, 0 message loss        |

Beyond component faults, the runbook covers the pipeline-shape invariant: a stage's output exceeding its input (duplication), a stage dropping to zero (blocked consumer), and DLQ growth with the [selective replay procedure](/orca/operations/dlq-replay).

The full runbook — symptoms, dashboards, step-by-step recovery procedures, escalation paths — is customer documentation, generated per deployment from your own chaos results and delivered at Gate 6 sign-off.

<Snippet file="cta-architect.mdx" />