Skip to main content

Operations

This section covers best practices for running your CSEs in production, including monitoring, recovery, and failover strategies.

Monitoring Health (Latency, Throughput, Errors)​

Each CSE includes built-in support for Prometheus / StatsD metrics export, allowing you to monitor system performance, health, and job statistics.

To enable Prometheus mode, configure the following environment variables:

METRICS_EXPORTER=prometheus
METRICS_PROMETHEUS_MODE=scrape
METRICS_PROMETHEUS_BIND_ADDR=0.0.0.0:4321

If you're running CSEs in Docker, don’t forget to bind the Prometheus port (e.g. 4321) to allow access from your metrics infrastructure.

We do not collect logs or metrics from your node. All exported metrics are for your internal monitoring only.

Metrics Sent to TACEO:Proof​

While we don’t ingest logs or metrics, your CSE does send minimal runtime measurements when submitting job results. These include:

  • Time to establish network connections
  • Time to run the coSNARK prover
  • Time to download blueprint artifacts (e.g., Proving Key, Verification Key, Circuit)

These are used solely for job-level tracing and auditing, not continuous telemetry.

Auto-Recovery & Failover​

Each CSE instance is stateless and identified by your secret key phrase. From TACEO:Proof’s perspective, restarting your CSE creates a fresh logical instance—there’s no persistent session management required.

If you’re using our official Docker image, it comes preconfigured with:

restart: unless-stopped

This ensures your CSE will automatically restart on crashes or restarts, keeping your node reliably online.

🛠️ Best Practice: On any unrecoverable error, your first line of defense is to restart the CSE process. This is safe and expected behavior in our architecture.