- Add Task.Supervisor for supervised background tasks in agent channel - Discovery worker retries transient failures (max 3 attempts) - Log silent polling timeouts with named task identification - Track errors in batch ARP/MAC upserts instead of silently ignoring - Discovery gracefully degrades on interface/sensor timeout (partial results) - Synchronous job dispatch on agent join to prevent race condition - Surface SNMP result, monitoring check, and backup processing errors - Atomic delete-stale + upsert for neighbors/ARP/MAC (transaction) - Discovery transaction timeout (60s) with duration logging - Heartbeat timeout detection (5min) closes stale agent channels - Live poll timeout (60s) notifies waiting LiveView on expiry - Disconnect agent channel when token is disabled or deleted - Debounce rapid assignment changes to prevent redundant job dispatches - Sensor index deduplication via YAML index template substitution |
||
|---|---|---|
| .. | ||
| device_oid.ex | ||
| device_profile.ex | ||
| mib_cache.ex | ||
| profile_watcher.ex | ||
| sensor_oid.ex | ||
| yaml_profiles.ex | ||