1. Foundation #
Getting started right.
Principles #
- Site chosen according to data location (EU for GDPR).
- Unified tagging
env,service,versioneverywhere, plusteam. - Agent deployed by the Kubernetes operator or configuration management, with a controlled version.
- API keys per use case, named and revocable application keys.
2. Costs #
The bill can be managed.
Levers #
| Item | Lever |
|---|---|
| Hosts and containers | Do not install the full agent where it is not needed |
| Logs | Filter in the agent, exclusion filters, short-retention indexes, archives |
| Custom metrics | No high-cardinality tags, Metrics without Limits |
| APM | Sampling and retention filters |
| Tracking | Usage page, alerts on usage metrics (datadog.estimated_usage.*) |
Good practice: Create monitors ondatadog.estimated_usage.logs.ingested_bytesanddatadog.estimated_usage.metrics.custom: a drift is detected within hours, not on the bill.
3. Alerts and governance #
Sustainable and shared.
Rules #
- Multi-alert monitors per service, with a runbook and a recipient team.
- SLOs for critical journeys, burn rate for on-call.
- Monitors, dashboards and SLOs in Terraform, reviewed in pull requests.
- OpenTelemetry instrumentation when possible, to keep freedom of choice.
Reference diagram and quick check #
Checklist.
Checklist #
- Site and keys configured
- Unified tagging verified
- Log pipelines and indexes under control
- APM with logs-traces correlation
- Monitors and SLOs in Terraform
- Usage alerts
