diff options
Diffstat (limited to 'Prometheus/alerts/lysergic/wireguard_exporters.yml')
| -rw-r--r-- | Prometheus/alerts/lysergic/wireguard_exporters.yml | 23 |
1 files changed, 23 insertions, 0 deletions
diff --git a/Prometheus/alerts/lysergic/wireguard_exporters.yml b/Prometheus/alerts/lysergic/wireguard_exporters.yml new file mode 100644 index 0000000..87fecb6 --- /dev/null +++ b/Prometheus/alerts/lysergic/wireguard_exporters.yml @@ -0,0 +1,23 @@ +groups: +- name: lysergic-wireguard_exporters + rules: + - alert: mercury-wireguard-down + expr: wireguard_peer_last_handshake_seconds{device="wg0", instance="localhost:9586", job="wireguard-mercury", public_key="$WG0_PUBKEY"} - time() <- 125 + for: 10s + labels: + job: wireguard-mercury + severity: CRITICAL + project: LYSERGIC + annotations: + title: WireGuard {{ $labels.instance }} is down + description: 'Failed to scrape {{ $labels.job }} on {{ $labels.instance}} for more than 2 minutes. Uplink to Psyched seems down. Affected services: Global infrastructure monitoring. Other outages might no longer be tracked. Urgent investigation is advised.' + - alert: lysergic-wireguard-down + expr: wireguard_peer_last_handshake_seconds{project="LYSERGIC",device!="wg1"} - time() <- 130 + for: 10s + labels: + job: wireguard-lysergic + severity: CRITICAL + project: LYSERGIC + annotations: + title: WireGuard Incident + description: 'No handshakes for {{ $labels.device }} in {{ $labels.job }} on {{ $labels.instance}} have been exchanged for an extended amount of time. Critical services could be affected.' |