Operational maturity
SLOs, capacity planning, disaster-recovery drills, environment promotion, and review cadence for Apache APISIX.
This page collects the operational maturity practices that keep an APISIX deployment reliable as it scales.
SLOs and alerting
| SLO | Signal | Example threshold | Source |
|---|---|---|---|
| Availability | apisix_http_status 5xx rate | < 0.1 % over 5 minutes | Prometheus |
| Latency | apisix_http_latency p99 | < 500 ms over 1 minute | Prometheus |
| Error budget | 5xx rate vs monthly target | burn rate > 14.4x | AlertManager |
| Configuration lag | etcd sync latency | < 1 second | APISIX control API |
- Export metrics from
/apisix/prometheus/metrics. - Route critical alerts to an on-call rotation. Record runbook links in every alert.
Capacity planning
- Establish baseline QPS and latency from performance tests.
- Model growth by 2x and 5x baselines.
- Plan CPU, memory, and etcd cluster size before reaching 70 % of current capacity.
- Re-evaluate after every major release and change in traffic mix.
Disaster recovery
- Keep
apisix/backuporetcdsnapshots in a separate region or account. - Document the recovery steps and test them quarterly.
- Maintain a cold-standby or multi-region plan if your RTO is less than one hour.
Environment promotion
- Development: validate plugins, routes, and CI/CD.
- Staging: mirror production traffic patterns and run load tests.
- Production: promote through a canary or blue-green process.
- Keep the same
config.yamlbase across environments; use environment-specific overrides.
Plugin metrics and troubleshooting
- Use the Plugin Common Configurations to set
_meta.error_response. - Troubleshooting covers common errors, etcd sync, TLS, and performance.
Content metadata
- Owner: Platform Engineering
- Last validated: TODO
- Review cadence: quarterly