ragleap-observability v0.3.0

Added

• AlertManager deployment (Deployment, Service, ConfigMap), wired end-to-end to Prometheus.
• First real alert rule: PostgresExporterDown, targeting a live, proven service.
• New ragleap-prometheus-rules ConfigMap mounted into Prometheus.

Fixed

• Prometheus's Deployment defaulted to RollingUpdate, but its ReadWriteOnce PVC holds a single-writer TSDB directory. The first real rollout crash-looped with lock DB directory: resource temporarily unavailable. Latent since the original deployment, exposed by this work. Fixed with strategy: Recreate.

Verified

• Prometheus /api/v1/rules confirms the rule loaded (health ok); /api/v1/alertmanagers confirms AlertManager registered as an active target; AlertManager /-/ready returns OK.

Known limitations

• No real notification receiver configured. Alerts are routed and grouped but not delivered to any human until a real Slack/email/PagerDuty receiver replaces the placeholder webhook.
• AlertManager state uses emptyDir (lost on restart).
• No canary or blue-green deployment strategy exists in the project; correctly sequenced after GitOps tooling (step 14).

Full details: CHANGELOG.md