diff options
Diffstat (limited to 'Prometheus/alerts/tripsit')
| -rw-r--r-- | Prometheus/alerts/tripsit/blackbox.yml | 21 | ||||
| -rw-r--r-- | Prometheus/alerts/tripsit/node_exporters.yml | 13 | 
2 files changed, 34 insertions, 0 deletions
diff --git a/Prometheus/alerts/tripsit/blackbox.yml b/Prometheus/alerts/tripsit/blackbox.yml new file mode 100644 index 0000000..0390718 --- /dev/null +++ b/Prometheus/alerts/tripsit/blackbox.yml @@ -0,0 +1,21 @@ +groups: +- name: tripsit-blackbox_exporters +  rules: +  - alert: TRIPSIT.ME-WEB-INCIDENT +    expr:  probe_success{instance=~"https://tripsit.me|https://chat.tripsit.me|https://chat.tripsit.me/chat|https://drugs.tripsit.me|https://benzo.tripsit.me|https://dxm.tripsit.me|https://combo.tripsit.me|https://wiki.tripsit.me|https://tripbot.tripsit.me"} == 0 +    for: 15s +    labels: +      severity: critical +      project: TRIPSIT +    annotations: +      title: Node {{ $labels.instance }} is down +      description: Failed to scrape {{ $labels.job }} on {{ $labels.instance}} for more than 15 seconds. Node seems down. +  - alert: TRIPSIT.DEV-WEB-INCIDENT +    expr: probe_success{instance=~"https://mail.tripsit.dev|https://dopamine.tripsit.dev"} == 0 +    for: 1m +    labels: +      severity: warning +      project: TRIPSIT +    annotations: +      title: Node {{ $labels.instance }} is down +      description: Failed to scrape {{ $labels.job }} on {{ $labels.instance}} for more than 1 minute. Node seems down. diff --git a/Prometheus/alerts/tripsit/node_exporters.yml b/Prometheus/alerts/tripsit/node_exporters.yml new file mode 100644 index 0000000..ee82367 --- /dev/null +++ b/Prometheus/alerts/tripsit/node_exporters.yml @@ -0,0 +1,13 @@ +groups: +- name: tripsit-node_exporters +  rules: +  - alert: TRIPSIT-NODE-INCIDENT +    expr: up{project="TRIPSIT", job="node_exporters"} == 0 +    for: 30s +    labels: +      severity: warning +      job: node_exporters +      project: TRIPSIT +    annotations: +      title: Node {{ $labels.instance }} is down +      description: Failed to scrape {{ $labels.job }} on {{ $labels.instance}} for more than 1 minute. Node seems down.  | 
