Monitoring

#Where to look

Surface	Access	Use
GET /v1/ops/status	Key with ops:read	JSON: overall status and one entry per check
GET /v1/ops/metrics	Key with ops:read	Prometheus text 0.0.4
pnpm cli check in apps/worker	Database access	The same checks; exits 1 when any check is critical

pnpm keys create monitoring --tenant <slug> --scopes ops:read

#Checks

Check	Fires when	First response
worker_heartbeat	No worker seen for 120 s	Is the worker running? Check database connectivity
mint_poll_freshness	Mint state not polled recently	Look for a job stuck on RPC retries
chain_clock_lag	Finalized clock over 120 s behind wall time	The RPC provider is stale
overdue_activations	A scheduled change is 300 s past due and unsettled	Look for a failing rebuild_timeline job for that mint
job_queue_draining	Pending jobs waiting over 30 min	Worker down or saturated
failed_jobs_24h	Jobs exhausted retries	Read jobs_outbox.last_error
balance_reconciliation	A fully replayed position differs from chain	Conversion is already disabled; treat it as a ledger bug
provider_agreement	The independent RPC disagrees (critical), or none is configured (warn)	See Independent RPC provider
failed_wallet_syncs	A wallet sync failed	Read wallet_syncs.error and re-request the sync
timeline_gaps	A mint's multiplier history has gaps	Affected positions are already partial
issuer_feed	Nothing loaded, or live data stale	With recorded fixtures this is ok by design

#Prometheus

scrape_configs:
  - job_name: corpact
    metrics_path: /v1/ops/metrics
    authorization: { type: Bearer, credentials_file: /etc/corpact/ops-key }
    static_configs: [{ targets: ['api.internal:4600'] }]

Alert on corpact_check_status == 2. The RPC budget comes from corpact_rpc_requests_total, and disagreements from corpact_provider_disagreements.