◆ PayGuardGitHub ↗
03 / OBSERVABILITY · SRE

See what the system knows.

A practical guide to the signals, rules, dashboards, and logs used to explain PayGuard health.

SIGNAL PATHS
METRICS

Spring Boot → Prometheus → Grafana

Micrometer exports HTTP and payment counters. A Go companion checks API health and exports its own probe metrics. Prometheus scrapes both endpoints and evaluates recording and alerting rules. Grafana queries the results.

LOGS

Containers → Alloy → Loki → Grafana

Alloy filters for PayGuard Compose containers, then forwards logs to Loki. Grafana queries them during failure investigation.

TERMS THAT MATTER

Measure behavior, not just uptime.

MetricA numeric measurement over time, such as request count or reconciliation attempts.
LogAn event record: a provider timeout or reconciliation result with context.
SLIA measured service indicator. PayGuard records five-minute availability and 250 ms latency ratios.
SLOA target for an SLI, used to decide when service behavior needs attention.
Error budgetThe allowed failure portion implied by the SLO. Burn rate describes how quickly it is spent.
AlertA Prometheus rule that evaluates when warning, critical burn rate, or latency conditions hold.
Readiness vs livenessReadiness controls traffic; liveness triggers a restart when the process cannot recover.
READ THE REAL CONFIG

Rules and dashboards.

View real Prometheus results and a Grafana dashboard capture ↗. GitHub Actions produces a dated snapshot daily, without relying on a laptop.

Recording rules include payguard:sli_availability:ratio5m, payguard:slo_availability:burn_rate5m and payguard:sli_latency_250ms:ratio5m. Alerts include warning and critical availability burn rate plus latency violation. Read the exact PromQL and thresholds ↗

Local dashboards: Grafana, Prometheus, Loki readiness. These links work only on a computer running PayGuard; they are not public hosted services.