Explore a provider timeout and its resolution. The scenario is an educational replay of behavior verified in GitHub Actions; the linked CI run is the executable proof.
A GitHub runner starts the real monitoring stack daily, queries Prometheus, and captures the provisioned Grafana dashboard. The result is hosted on GitHub Pages with a UTC timestamp, so visitors can view it without your laptop.
This is a dated capture. The runner shuts down after publishing; it does not expose a continuously running Prometheus server.
The client POSTs a payment with an Idempotency-Key. That key identifies this logical payment across retries.
POST /api/v1/payments
Idempotency-Key: DEMO-TIMEOUT-...
{ "customerId": "CUSTOMER-DEMO", "amount": 149900, "currency": "INR" }The key belongs to the payment attempt, not to a single HTTP connection.
These controls replay the documented behavior locally in your browser. They do not send a payment or query a live production API.
Prometheus records a five-minute ratio and compares it with a 99.9% availability objective.
payguard:sli_availability:ratio5mThe burn rate relates observed error ratio to the allowed 0.1% error budget. Warning and critical rules evaluate it.
payguard:slo_availability:burn_rate5mA five-minute ratio tracks responses within 250 ms; the latency alert evaluates a 99% threshold.
payguard:sli_latency_250ms:ratio5mAlloy ships scoped PayGuard logs to Loki. Provider and reconciliation events help explain the UNKNOWN → SUCCESS transition.
provider_timeout_after_successLearn the monitoring terms → · Inspect actual Prometheus rules ↗
See the latest captured Prometheus queries and Grafana dashboard ↗. The GitHub runner refreshes this dated evidence daily.
The badge shows the latest workflow outcome. The run page contains a summary and a downloadable evidence artifact.
Open recent runs ↗Real metric charts and container logs are available on the machine running Docker Compose.
Open overview locally ↗Inspect the actual health response and Prometheus targets.
API health ↗The public repository explains how the worker, fake provider, database, and rules produce these results.
Request flow →Public visitors can use the interactive replay and hosted CI results. localhost links refer to each visitor’s own computer and require PayGuard to be running there.