High Availability
Technical guidance for High Availability in Grafana, with production, operation and evidence-driven troubleshooting context.
Operational point 1
Operational note 1 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 2
Operational note 2 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 3
Operational note 3 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 4
Operational note 4 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 5
Operational note 5 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 6
Operational note 6 for High Availability: validate this point against the documentation for the exact installed version before applying production changes.
Operational point 7
Grafana 13 update: application HA and Alerting HA are separate layers. Grafana uses a shared MySQL/PostgreSQL database for HA; Alerting has its own coordination, with Memberlist preferred and Redis available when direct node communication is not possible. Single-node evaluation reduces duplicate queries but introduces a brief failover gap.
Engineering workflow
Scope and impact
Define service, owner, expected state and impact before changing the monitor.
Evidence first
Validate source, path, credentials, baseline and timestamps before changing thresholds.
Controlled change
Change one variable at a time and retain a rollback path.
Commands / reference flow
Cliente -> Load Balancer -> Grafana A --\
Grafana B ----> MySQL/PostgreSQL compartilhadoRecommended operating routine
- Review critical states, collection gaps and delivery failures every shift.
- Review noisy alerts, slow objects, credentials and recurring errors every week.
- Review versions, backup/restore, permissions and capacity periodically.
Symptom → scope → layer → evidence → hypothesis → validation
Build, break, diagnose and recover
- Record a healthy baseline.
- Induce one controlled failure related to this topic.
- Identify the failing layer before restarting components.
- Correct the identified cause and validate recovery.
Ready for production?
Ownership
Objective, scope and owner are defined.
Security
Credentials and integrations follow least privilege.
Validation
Failure and recovery were tested and documented.
Version-sensitive documentation
Editorial reference reviewed against Grafana 13.2. Reconfirm APIs, breaking changes, plugins, and HA behavior against the exact installed release.