Monitoring & alerts

What to watch

Three signals matter for a healthy Curio Core SP:

1. Chain head is advancing. If Lantern's head stops ticking, every downstream task stalls. Dashboard Overview shows it live.
2. Proof loop is succeeding. 24h prove succeed/fail counts on the Overview card. Steady-state should be ~12 successes per dataset per day with zero fails.
3. PDP wallet has gas. Below ~0.1 tFIL/FIL, proofs start failing on insufficient funds. Wallets page shows live balance.

Everything else (task queue depth, rail count, alert population) is secondary.

Alerts

The alerts table accumulates rows from a 30-second poller task. Today, alert triggers are sparse — only the alert manager's own self-checks (rate limit on the poller, etc.) populate it. The dashboard's Alerts page is in place and ready; the trigger surface is being built out iteration by iteration.

Planned alert triggers:

• Low PDP wallet balance (< threshold)
• Missed prove cycle (consecutive_prove_failures > 2)
• Stuck task (unowned + non-eligible for budget reasons)
• Rail terminated by FWSS

Logs

sudo journalctl -u curio-core -f

Useful filters:

# Just the prove loop
sudo journalctl -u curio-core -f | grep -E "PDPv0_Prove|provePossession"

# Just the payment settler
sudo journalctl -u curio-core -f | grep -E "payments|settleRail"

# Errors only
sudo journalctl -u curio-core -f | grep -E "ERROR|WARN"

External monitoring

Curio Core doesn't yet ship Prometheus / OpenTelemetry exporters. For external monitoring, periodically scrape:

GET /api/overview

Returns JSON with chain head, dataset count, piece count, rail count, 24h prove succeed/fail, scheduler health. Cheap to call (SQL only, no chain reads).

For Lantern-side telemetry, see the Lantern docs.