Monitoring Vault State

Combine state polling with notification channels. Notifications deliver alerts; they are not a complete execution ledger. These upstream paths require the applicable access permissions; agree polling intervals with the integration owner.

​

Key Metrics

​

PPS Staleness

Vault detail exposes pps_info.last_pps_update_timestamp (Unix seconds) and pps_info.max_pps_update_interval (seconds). Compare their sum with current time to determine remaining PPS lifetime. Alert as PPS approaches or reaches expiry using the vault’s operating policy. A missing value, invalid timestamp, or unavailable observation freshness is UNKNOWN, not extra lifetime. Check strategy_is_paused separately; an absent value does not mean unpaused.

​

Service Success Rate

The services endpoint returns control records such as service_name, is_paused, run_interval_seconds, last_run_at, and next_run_at, not a success_rate. Inspect pause state and run timing. Calculate a success rate only from a verified run-history source with a defined window and sample count; no samples means UNKNOWN, not 100% success.

​

Redemption Queue Depth

Track pending request count, value, and age from redemption data. Do not count generic pending executions as redemption requests. Reconcile fulfillment and claimable assets separately; see Redemption.

​

Full Monitoring Loop

1. Fetch through an approved integration with timeouts, bounded backoff, and no overlapping polls.
2. Validate the response envelope and required fields. Retain collection time separately from producer observation time.
3. Classify missing, stale, malformed, or non-finite values and missing or future observation times as UNKNOWN. A successful HTTP response alone does not establish freshness or health.
4. Apply thresholds only to valid, sufficiently fresh observations. Persist the result and alert on both degraded and unknown conditions.
5. Investigate the affected vault, service, or intent before any separately authorized remediation.

​

Alert Integrations

Route alerts through configured notification channels or your approved incident integration. Include vault identity, condition, observation time, and relevant intent or transaction IDs. Deduplicate repeated alerts and keep destination secrets server-side.

​

Slack

Use the approved channel destination and include a link to the affected vault or intent for investigation.

​

PagerDuty

Map conditions to your incident policy, keep a stable incident key, and resolve only after a valid recovery observation.