diff --git a/alerts/alert8_network_hardware_alarm.json b/alerts/alert8_network_hardware_alarm.json new file mode 100644 index 0000000..422b76d --- /dev/null +++ b/alerts/alert8_network_hardware_alarm.json @@ -0,0 +1,21 @@ +{ + "title": "CRITICAL: network equipment hardware/power alarm", + "description": "Any-match critical alarm from the new BDCOM EPON/GPON, ZTE, Juniper, or D-Link rules (rule27-84): ONU dying gasp/power loss, ONU config/HAL driver failure, PSU failure, memory parity error, ASIC/PFE error, or a switch cold start. Same 'fires on every occurrence, no threshold' philosophy as alert4_juniper_chassis_alarm - these are hardware-level failures, not transient noise. Built from the user's log-signature report; not yet confirmed against real device output (see rules/rule27-84 descriptions).", + "priority": 3, + "alert": true, + "config": { + "type": "aggregation-v1", + "query": "event_type:dying_gasp OR event_type:onu_config_error OR event_type:hal_driver_error OR event_type:onu_info_failed OR event_type:psu_failure OR event_type:memory_parity_error OR event_type:asic_error OR event_type:system_reboot", + "streams": ["__NETWORK_STREAM_ID__"], + "group_by": [], + "series": [], + "conditions": {"expression": null}, + "search_within_ms": 60000, + "execute_every_ms": 60000, + "event_limit": 100 + }, + "field_spec": {}, + "key_spec": [], + "notification_settings": {"grace_period_ms": 600000, "backlog_size": 5}, + "notifications": [{"notification_id": "__DISCORD_NOTIFICATION_ID__"}] +} diff --git a/alerts/alert9_network_repeated_warning.json b/alerts/alert9_network_repeated_warning.json new file mode 100644 index 0000000..a447be7 --- /dev/null +++ b/alerts/alert9_network_repeated_warning.json @@ -0,0 +1,21 @@ +{ + "title": "WARNING: repeated network equipment warning from one device", + "description": "Same 'single occurrence can be transient, repetition means a real problem' philosophy as alert7_accelppp_repeated_error_warn, applied to the new Network Equipment rules (rule27-84): repeated ONU signal-loss/offline, optical low power, BGP session resets, LACP timeouts, link flapping, DDoS violations, SNMP auth violations, etc. from the same device. Threshold matches alert7 (>5 in 5 minutes). Built from the user's log-signature report; not yet confirmed against real device output.", + "priority": 2, + "alert": true, + "config": { + "type": "aggregation-v1", + "query": "severity_tag:warning", + "streams": ["__NETWORK_STREAM_ID__"], + "group_by": ["gl2_remote_ip", "event_type"], + "series": [{"type": "count", "id": "count-", "field": null}], + "conditions": {"expression": {"expr": ">", "left": {"expr": "number-ref", "ref": "count-"}, "right": {"expr": "number", "value": 5.0}}}, + "search_within_ms": 300000, + "execute_every_ms": 300000, + "event_limit": 50 + }, + "field_spec": {}, + "key_spec": [], + "notification_settings": {"grace_period_ms": 900000, "backlog_size": 5}, + "notifications": [{"notification_id": "__DISCORD_NOTIFICATION_ID__"}] +}