Runbooks¶
Operational playbooks for responding to incidents and routine maintenance on the CSA-in-a-Box platform. Each runbook is structured for first-responder use: detection signals, triage steps, mitigation, root-cause analysis, and links to the relevant ADRs and code.
Incident response¶
| Runbook | When to use |
|---|---|
| Break-Glass Access | Emergency elevation when normal access paths are unavailable |
| Security Incident | Suspected breach, credential exposure, or anomalous activity |
| Data Pipeline Failure | ADF/Databricks pipeline failures, retry exhaustion, dead-letter overflow |
| Dead Letter Queue | Drop-event triage from streaming and batch pipelines |
| DR Drill | Quarterly disaster-recovery exercises and the live failover playbook |
| Purview Scan Failure | Catalog scans not completing or returning unexpected results |
| Azure OpenAI Throttling | 429s from Azure OpenAI; capacity reallocation |
| Databricks Cost Runaway | Spend anomaly response for Databricks clusters |
| Cost Alert Response | Budget threshold response across the platform |
Routine operations¶
| Runbook | When to use |
|---|---|
| Certificate Expiration | Pre-emptive cert rotation and last-minute renewal |
| Key Rotation | Scheduled rotation of Key Vault keys and secrets |
| Azure Deployment Service Principal | Managing the limitlessdata_deploy SP used by CI/CD |
| Copilot Studio Agent Rotation | Refreshing Copilot Studio agent registrations |
| Fabric Capacity Management | Scaling Fabric SKUs up/down based on workload |
| Tenant Onboarding | Adding a new tenant to multi-tenant landing zones |
| Portal Rollout Strategy | Staged portal release process |
| dbt CI | dbt pipeline CI workflow, hotfixes, and rerun playbook |
| Release-Please Status Bypass | Working around release-please CI status when it stalls |
Related¶
- Disaster Recovery — strategy and RTO/RPO targets
- Multi-Region — failover topology
- Production Checklist — pre-cutover validation
- Troubleshooting — diagnostic flowcharts