Configuration for monitoring APC PDU

3 years ago · c5e6fbcfdf
parent 91fe213e02
commit c5e6fbcfdf
7 changed files with 506 additions and 3 deletions
--- a/monitoring.yml
+++ b/monitoring.yml
@ -58,6 +58,8 @@
    prometheus_alertmanager: docker-ovh.adm.auro.re:9093
    snmp_unifi_password: "{{ vault_snmp_unifi_password }}"
    snmp_switch_community: "{{ vault_snmp_switch_community }}"
+    snmp_pdu_user: "{{ vault_snmp_pdu_user }}"
+    snmp_pdu_password: "{{ vault_snmp_pdu_password }}"

    # Prometheus targets.json
    prometheus_targets:
@ -67,6 +69,7 @@
      - targets: "{{ groups['gs_unifi'] | list | sort }}"
    prometheus_ups_snmp_targets:
      - ups-gk-1.ups.auro.re
+    prometheus_pdu_snmp_targets:
      - pdu-ga-1.ups.auro.re
  roles:
    - prometheus
--- a/roles/prometheus/tasks/main.yml
+++ b/roles/prometheus/tasks/main.yml
@ -50,14 +50,16 @@
    line: "ARGS=\"--web.listen-address=127.0.0.1:9116\""
  notify: Restart prometheus-snmp-exporter

-# This file store SNMP OIDs
+# These files store SNMP OIDs
 - name: Configure Prometheus snmp-exporter
  template:
-    src: snmp.yml.j2
-    dest: /etc/prometheus/snmp.yml
+    src: "{{ item }}.j2"
+    dest: "/etc/prometheus/{{ item }}"
    owner: prometheus
    group: prometheus
    mode: u=r,g=r,o=
+  loop:
+    - snmp.yml
  notify: Restart prometheus-snmp-exporter

 # We don't need to restart Prometheus when updating nodes
@ -103,6 +105,13 @@
    mode: 0644
  when: prometheus_postgres_targets is defined

+- name: Configure Prometheus apc_pdu monitoring
+  copy:
+    content: "{{ [{'targets': prometheus_pdu_snmp_targets }] | to_nice_json }}\n"
+    dest: /etc/prometheus/targets_apc_pdu_snmp.json
+    mode: 0644
+  when: prometheus_pdu_snmp_targets is defined
+
 - name: Activate prometheus service
  systemd:
    name: prometheus
--- a/roles/prometheus/templates/prometheus.yml.j2
+++ b/roles/prometheus/templates/prometheus.yml.j2
@ -123,4 +123,21 @@ scrape_configs:
      - source_labels: [__param_target]
        target_label: __address__
        replacement: '$1:9187'
+
+  - job_name: apc_pdu_snmp
+    file_sd_configs:
+      - files:
+        - '/etc/prometheus/targets_apc_pdu_snmp.json'
+    metrics_path: /snmp
+    params:
+      module:
+      - apc_pdu
+    relabel_configs:
+      - source_labels: [__address__]
+        target_label: __param_target
+      - source_labels: [__param_target]
+        target_label: instance
+      - target_label: __address__
+        replacement: 127.0.0.1:9116
+
 ...
--- a/roles/prometheus/templates/snmp.yml.j2
+++ b/roles/prometheus/templates/snmp.yml.j2
@ -416,4 +416,70 @@ ubiquiti_unifi:
    auth_protocol: SHA
    priv_protocol: AES
    priv_password: {{ snmp_unifi_password }}
+
+
+apc_pdu:
+  walk:
+  - 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.5
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+  - 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+  - 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+  metrics:
+  - name: rPDU2SensorTempHumidityStatusTempC
+    oid: 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+    type: gauge
+    help: Sensor temperature reading in tenths of degrees Celsius - 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+    indexes:
+    - labelname: rPDU2SensorTempHumidityStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusLoadState
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+    type: gauge
+    help: Indicates the present load status of the Rack PDU - 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusPower
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.5
+    type: gauge
+    help: The power consumption of the Rack PDU load in hundredths of kilowatts -
+      1.3.6.1.4.1.318.1.1.26.4.3.1.5
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusPeakPower
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+    type: gauge
+    help: The peak power consumption of the Rack PDU load in hundredths of kilowatts
+      - 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2PhaseStatusPowerFactor
+    oid: 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+    type: gauge
+    help: Indicates the load power factor, in hundredths, of the Rack PDU phase being
+      queried - 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+    indexes:
+    - labelname: rPDU2PhaseStatusIndex
+      type: gauge
+  - name: rPDU2OutletMeteredStatusPower
+    oid: 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+    type: gauge
+    help: Indicates the power draw of the load on the Rack PDU outlet being queried
+      - 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+    indexes:
+    - labelname: rPDU2OutletMeteredStatusIndex
+      type: gauge
+  version: 3
+  auth:
+    security_level: authPriv
+    username: {{ snmp_pdu_user }}
+    password: {{ snmp_pdu_password }}
+    auth_protocol: SHA
+    priv_protocol: AES
+    priv_password: {{ snmp_pdu_password }}
+
 ...
--- a/roles/prometheus/templates/snmp_apc_pdu.yml.j2
+++ b/roles/prometheus/templates/snmp_apc_pdu.yml.j2
@ -0,0 +1,68 @@
+---
+{{ ansible_managed | comment }}
+
+apc_pdu:
+  walk:
+  - 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.5
+  - 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+  - 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+  - 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+  metrics:
+  - name: rPDU2SensorTempHumidityStatusTempC
+    oid: 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+    type: gauge
+    help: Sensor temperature reading in tenths of degrees Celsius - 1.3.6.1.4.1.318.1.1.26.10.2.2.1.8
+    indexes:
+    - labelname: rPDU2SensorTempHumidityStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusLoadState
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+    type: gauge
+    help: Indicates the present load status of the Rack PDU - 1.3.6.1.4.1.318.1.1.26.4.3.1.4
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusPower
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.5
+    type: gauge
+    help: The power consumption of the Rack PDU load in hundredths of kilowatts -
+      1.3.6.1.4.1.318.1.1.26.4.3.1.5
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2DeviceStatusPeakPower
+    oid: 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+    type: gauge
+    help: The peak power consumption of the Rack PDU load in hundredths of kilowatts
+      - 1.3.6.1.4.1.318.1.1.26.4.3.1.6
+    indexes:
+    - labelname: rPDU2DeviceStatusIndex
+      type: gauge
+  - name: rPDU2PhaseStatusPowerFactor
+    oid: 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+    type: gauge
+    help: Indicates the load power factor, in hundredths, of the Rack PDU phase being
+      queried - 1.3.6.1.4.1.318.1.1.26.6.3.1.9
+    indexes:
+    - labelname: rPDU2PhaseStatusIndex
+      type: gauge
+  - name: rPDU2OutletMeteredStatusPower
+    oid: 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+    type: gauge
+    help: Indicates the power draw of the load on the Rack PDU outlet being queried
+      - 1.3.6.1.4.1.318.1.1.26.9.4.3.1.7
+    indexes:
+    - labelname: rPDU2OutletMeteredStatusIndex
+      type: gauge
+  version: 3
+  auth:
+    security_level: authPriv
+    username: {{ snmp_pdu_user }}
+    password: {{ snmp_pdu_password }}
+    auth_protocol: SHA
+    priv_protocol: AES
+    priv_password: {{ snmp_pdu_password }}
+
+...
--- a/roles/prometheus/templates/snmp_procurve_switcs.yml.j2
+++ b/roles/prometheus/templates/snmp_procurve_switcs.yml.j2
@ -0,0 +1,72 @@
+---
+{{ ansible_managed | comment }}
+
+procurve_switch:
+  walk:
+    - 1.3.6.1.2.1.31.1.1.1.10
+    - 1.3.6.1.2.1.31.1.1.1
+    - 1.3.6.1.2.1.2.2.1.2
+    - 1.3.6.1.2.1.31.1.1.1.18
+  get:
+    - 1.3.6.1.2.1.1.3.0
+    - 1.3.6.1.2.1.1.5.0
+    - 1.3.6.1.2.1.1.6.0
+  metrics:
+    - name: sysUpTime
+      oid: 1.3.6.1.2.1.1.3
+      type: gauge
+      help: The time (in hundredths of a second) since the network management
+        portion of the system was last re-initialized. - 1.3.6.1.2.1.1.3
+    - name: sysName
+      oid: 1.3.6.1.2.1.1.5
+      type: DisplayString
+      help: An administratively-assigned name for this managed node
+        - 1.3.6.1.2.1.1.5
+    - name: sysLocation
+      oid: 1.3.6.1.2.1.1.6
+      type: DisplayString
+      help: The physical location of this node (e.g., 'telephone closet, 3rd
+        floor') - 1.3.6.1.2.1.1.6
+    - name: ifHCOutOctets
+      oid: 1.3.6.1.2.1.31.1.1.1.10
+      type: counter
+      help: The total number of octets transmitted out of the interface,
+        including framing characters - 1.3.6.1.2.1.31.1.1.1.10
+      indexes:
+        - labelname: ifIndex
+          type: gauge
+      lookups:
+        - labels:
+            - ifIndex
+          labelname: ifDescr
+          oid: 1.3.6.1.2.1.2.2.1.2
+          type: DisplayString
+        - labels:
+            - ifIndex
+          labelname: ifName
+          oid: 1.3.6.1.2.1.31.1.1.1.1
+          type: DisplayString
+    - name: ifHCInOctets
+      oid: 1.3.6.1.2.1.31.1.1.1.6
+      type: counter
+      help: The total number of octets received on the interface, including
+        framing characters - 1.3.6.1.2.1.31.1.1.1.6
+      indexes:
+        - labelname: ifIndex
+          type: gauge
+      lookups:
+        - labels:
+            - ifIndex
+          labelname: ifDescr
+          oid: 1.3.6.1.2.1.2.2.1.2
+          type: DisplayString
+        - labels:
+            - ifIndex
+          labelname: ifName
+          oid: 1.3.6.1.2.1.31.1.1.1.1
+          type: DisplayString
+  version: 2
+  auth:
+    community: "{{ snmp_switch_community }}"
+
+...
--- a/roles/prometheus/templates/snmp_ubiquiti_unifi.yml.j2
+++ b/roles/prometheus/templates/snmp_ubiquiti_unifi.yml.j2
@ -0,0 +1,268 @@
+---
+{{ ansible_managed | comment }}
+
+ubiquiti_unifi:
+  walk:
+  - 1.3.6.1.4.1.41112.1.6
+  get:
+  - 1.3.6.1.2.1.1.5.0
+  - 1.3.6.1.2.1.1.6.0
+  metrics:
+# Pour faire une WifiMap un jour, on peut entrer la location dans la conf des bornes
+#  - name: sysLocation
+#    oid: 1.3.6.1.2.1.1.6
+#    type: DisplayString
+#    help: The physical location of this node (e.g., 'telephone closet, 3rd floor')
+#      - 1.3.6.1.2.1.1.6
+  - name: unifiVapIndex
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.1
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.1'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapChannel
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.4
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.4'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapEssId
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.6
+    type: DisplayString
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.6'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapName
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.7
+    type: DisplayString
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.7'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifi_vap_num_stations
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.8
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.8'
+    indexes:
+    - labelname: unifi_vap_index
+      type: gauge
+    lookups:
+    - labels: [unifi_vap_index]
+      labelname: unifi_vap_essid
+      oid: 1.3.6.1.4.1.41112.1.6.1.2.1.6
+      type: DisplayString
+    - labels: [unifi_vap_index]
+      labelname: unifi_vap_radio
+      oid: 1.3.6.1.4.1.41112.1.6.1.2.1.9
+      type: DisplayString
+    - labels: []
+      labelname: unifi_vap_index
+#  - name: unifiVapNumStations
+#    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.8
+#    type: gauge
+#    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.8'
+#    indexes:
+#    - labelname: unifiVapIndex
+#      type: gauge
+  - name: unifiVapRadio
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.9
+    type: DisplayString
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.9'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxBytes
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.10
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.10'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxCrypts
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.11
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.11'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxDropped
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.12
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.12'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxErrors
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.13
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.13'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxFrags
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.14
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.14'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapRxPackets
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.15
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.15'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxBytes
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.16
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.16'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxDropped
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.17
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.17'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxErrors
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.18
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.18'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxPackets
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.19
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.19'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxRetries
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.20
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.20'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapTxPower
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.21
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.21'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapUp
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.22
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.1.2.1.22'
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiVapUsage
+    oid: 1.3.6.1.4.1.41112.1.6.1.2.1.23
+    type: DisplayString
+    help: guest or regular user - 1.3.6.1.4.1.41112.1.6.1.2.1.23
+    indexes:
+    - labelname: unifiVapIndex
+      type: gauge
+  - name: unifiIfIndex
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.1
+    type: gauge
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.1'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfName
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.5
+    type: DisplayString
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.5'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfRxBytes
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.6
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.6'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfRxDropped
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.7
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.7'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfRxError
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.8
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.8'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfRxMulticast
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.9
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.9'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfRxPackets
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.10
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.10'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfTxBytes
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.12
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.12'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfTxDropped
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.13
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.13'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfTxError
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.14
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.14'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiIfTxPackets
+    oid: 1.3.6.1.4.1.41112.1.6.2.1.1.15
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.2.1.1.15'
+    indexes:
+    - labelname: unifiIfIndex
+      type: gauge
+  - name: unifiApSystemModel
+    oid: 1.3.6.1.4.1.41112.1.6.3.3
+    type: DisplayString
+    help: ' - 1.3.6.1.4.1.41112.1.6.3.3'
+  - name: unifiApSystemUptime
+    oid: 1.3.6.1.4.1.41112.1.6.3.5
+    type: counter
+    help: ' - 1.3.6.1.4.1.41112.1.6.3.5'
+  version: 3
+  auth:
+    security_level: authPriv
+    username: snmp_prometheus
+    password: {{ snmp_unifi_password }}
+    auth_protocol: SHA
+    priv_protocol: AES
+    priv_password: {{ snmp_unifi_password }}
+...