Increase alerting threshold for CPU pressure

[chef.git] / cookbooks / prometheus / templates / default / alert_rules.yml.erb

diff --git a/cookbooks/prometheus/templates/default/alert_rules.yml.erb b/cookbooks/prometheus/templates/default/alert_rules.yml.erb
index 562d31046fbb2a610c4f25101b12d2f15b5d5d3c..0aa0b64cbea271981d34dfaec55bdef89f5118d4 100644 (file)
--- a/cookbooks/prometheus/templates/default/alert_rules.yml.erb
+++ b/cookbooks/prometheus/templates/default/alert_rules.yml.erb
@@ -22,10 +22,19 @@ groups:
           alertgroup: "{{ $labels.instance }}"
         annotations:
           busy_workers: "{{ $value | humanizePercentage }}"
+  - name: chef
+    rules:
+      - alert: chef client not running
+        expr: time() - node_systemd_timer_last_trigger_seconds{name="chef-client.timer"} > 3600
+        for: 12h
+        labels:
+          alertgroup: "{{ $labels.instance }}"
+        annotations:
+          down_time: "{{ $value | humanizeDuration }}"
   - name: cpu
     rules:
       - alert: cpu pressure
-        expr: rate(node_pressure_cpu_waiting_seconds_total[5m]) > 0.3
+        expr: rate(node_pressure_cpu_waiting_seconds_total[5m]) > 0.6
         for: 15m
         labels:
           alertgroup: "{{ $labels.instance }}"
@@ -40,6 +49,15 @@ groups:
           alertgroup: database
         annotations:
           delay: "{{ $value | humanizeDuration }}"
+  - name: fastly
+    rules:
+      - alert: error rate
+        expr: sum(rate(fastly_rt_status_group_total{status_group="5xx"}[5m])) by (service_name, datacenter) / sum(rate(fastly_rt_status_group_total[5m])) by (service_name, datacenter) > 0.005
+        for: 15m
+        labels:
+          alertgroup: fastly
+        annotations:
+          error_rate: "{{ $value | humanizePercentage }}"
   - name: filesystem
     rules:
       - alert: readonly filesystem
@@ -358,10 +376,15 @@ groups:
   - name: systemd
     rules:
       - alert: systemd failed service
-        expr: node_systemd_unit_state{state="failed"} == 1
+        expr: node_systemd_unit_state{state="failed",name!="chef-client.service"} == 1
         for: 5m
         labels:
           alertgroup: "{{ $labels.instance }}"
+      - alert: systemd failed service
+        expr: node_systemd_unit_state{state="failed",name="chef-client.service"} == 1
+        for: 6h
+        labels:
+          alertgroup: "{{ $labels.instance }}"
   - name: tile
     rules:
       - alert: renderd replication delay
@@ -401,3 +424,10 @@ groups:
           alertgroup: web
         annotations:
           error_rate: "{{ $value | humanizePercentage }}"
+      - alert: job processing rate
+        expr: rate(pg_stat_user_tables_n_tup_del{datname="openstreetmap",relname="delayed_jobs"}[5m]) / rate(pg_stat_user_tables_n_tup_ins{datname="openstreetmap",relname="delayed_jobs"}[5m]) < 0.9 and ignoring(job, name, datname, relname, schemaname, server) chef_role{name="db-master"} == 1
+        for: 5m
+        labels:
+          alertgroup: web
+        annotations:
+          job_processing_rate: "{{ $value | humanizePercentage }}"