Release candidate: this page describes 410fbfb, which is separate from main. See version and availability.

Run your first monitor¶

This walkthrough uses a local HTTP target and a log notification. You need Git, Go 1.25 or later, Make, and Python 3 for the demo target. The dashboard assets are already included in the Go source.

1. Build CPRa¶

git clone --branch codex/release-engineering https://github.com/ziad-hsn/cpra.git
cd cpra
git checkout --detach 410fbfb0092d01277b3884cd04151c27443a4226
make
./bin/cpra -version

The build produces bin/cpra and bin/cpractl. See installation for optional drivers and container builds.

2. Start a local target¶

In another terminal:

mkdir -p /tmp/cpra-demo
printf 'ok\n' > /tmp/cpra-demo/health
python3 -m http.server 8080 --bind 127.0.0.1 --directory /tmp/cpra-demo

This serves http://127.0.0.1:8080/health.

3. Start CPRa¶

In the CPRa checkout:

cp examples/monitors.yaml monitors.yaml
./bin/cpra -yaml monitors.yaml -data-dir ./demo-state

The example checks the target every 30 seconds, uses a five-second timeout, and writes incident transitions to alerts.jsonl. Keep both processes running. This demo explicitly stores durable state under ./demo-state; retain that directory if you restart the same demo.

4. Inspect the monitor¶

Open the local dashboard, or use another terminal:

./bin/cpractl health
./bin/cpractl get monitors
./bin/cpractl get overview -o json

A new monitor can appear as unknown before its first check. Allow a check interval and a dashboard snapshot refresh before assessing its state.

5. Observe a failure and recovery¶

Stop the demo HTTP server. After the configured consecutive failures, inspect the dashboard and alerts.jsonl. Restart the same demo server and allow the configured consecutive healthy checks for recovery.

This example has no recovery action configured: you restart the target yourself. Add an action deliberately after confirming the monitoring and notification behavior.

Use Ctrl+C to stop CPRa and the demo target. The default listener is local; use the authenticated deployment guide for remote access.

Use cpractl get state to inspect storage and cpractl get history MONITOR_ID for the timeline. Obtain the stable ID from the monitor JSON response. Stopping the demo leaves its manifest, alert log and state directory in place; it does not delete retained history.