Merge pull request 'Merge monitoring of new ups, and adapt threesold' (#33) from monitoring_ups into master

Reviewed-on: Aurore/ansible#33
Reviewed-by: otthorn <otthorn@noreply.localhost>
Reviewed-by: jeltz <jeltz@noreply.localhost>

hosts

@@ -31,7 +31,6 @@ re2o-ldap.adm.auro.re
 re2o-db.adm.auro.re
 #services-bdd-local.adm.auro.re
 backup.adm.auro.re
-services-web.adm.auro.re
 mail.adm.auro.re
 wikijs.adm.auro.re
 prometheus-aurore.adm.auro.re

monitoring.yml

@@ -4,6 +4,7 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_targets:
@@ -18,6 +19,7 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_targets:
@@ -27,6 +29,7 @@
       - targets: "{{ groups['pacaterie_unifi'] | list | sort }}"
     prometheus_ups_snmp_targets:
       - ups-pn-1.ups.auro.re
+      - ups-ps-1.ups.auro.re
   roles:
     - prometheus
 
@@ -34,10 +37,12 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_ups_snmp_targets:
       - ups-ec-1.ups.auro.re
+      - ups-ec-2.ups.auro.re
 
     prometheus_targets:
       - targets: |
@@ -51,6 +56,7 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_targets:
@@ -67,6 +73,7 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_ups_snmp_targets:
@@ -100,6 +107,7 @@
   vars:
     prometheus_alertmanager: docker-ovh.adm.auro.re:9093
     snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
+    snmp_switch_community: "{{ vault_snmp_switch_community }}"
 
     # Prometheus targets.json
     prometheus_targets:

roles/prometheus/templates/prometheus/alert.rules.yml.j2

@@ -119,7 +119,7 @@ groups:
       summary: "La tension de sortie de {{ $labels.instance }} est de {{ $value }}V."  
 
   - alert: UpsTimeRemainingWarning
-    expr: upsEstimatedMinutesRemaining < 15
+    expr: upsEstimatedMinutesRemaining < 8
     for: 1m
     labels:
       severity: warning

roles/prometheus/templates/prometheus/snmp.yml.j2

@@ -14,6 +14,7 @@ eatonups:
   - 1.3.6.1.2.1.33.1.3
   - 1.3.6.1.2.1.33.1.4
   - 1.3.6.1.4.1.534.1.6
+  - 1.3.6.1.4.1.318.1.1.10.2.3.2.1.4
   get:
   - 1.3.6.1.2.1.1.3.0
   metrics:
@@ -68,9 +69,10 @@ eatonups:
     - labelname: upsOutputLineIndex
       type: gauge
   - name: xupsEnvRemoteTemp
-    oid: 1.3.6.1.4.1.534.1.6.5
+#    oid: 1.3.6.1.4.1.534.1.6.5
+    oid: 1.3.6.1.4.1.318.1.1.10.2.3.2.1.4
     type: gauge
-    help: The reading of an EMP's temperature sensor. - 1.3.6.1.4.1.534.1.6.5
+    help: The reading of an EMP's temperature sensor. - 1.3.6.1.4.1.318.1.1.10.2.3.2.1.4
   - name: xupsEnvRemoteHumidity
     oid: 1.3.6.1.4.1.534.1.6.6
     type: gauge