Add checks for the R720's thermal monitors, also add those thermal monitors

This commit is contained in:
Salt 2022-01-10 22:14:07 -06:00
parent 8a3f13a939
commit 8517e842b2
5 changed files with 23 additions and 3 deletions

3
.gitmodules vendored Normal file
View File

@ -0,0 +1,3 @@
[submodule ".role-submodules/ansible-role-lm-sensors"]
path = roles/ansible-role-lm-sensors
url = https://github.com/aisbergg/ansible-role-lm-sensors

View File

@ -0,0 +1,9 @@
#!/usr/bin/env ansible-playbook
# vim:ft=ansible:
---
- hosts: device_types_r720
roles:
- role: ansible-role-lm-sensors
vars:
lm_sensors_force_detection: yes
tags: [ lm-sensors ]

View File

@ -218,11 +218,17 @@
hostgroup: "ansible,!role-hypervisor"
- name: Users
command: check_by_ssh!/usr/lib/nagios/plugins/check_users -w 3 -c 5
# Privileged checks
# Required because check_disk may attempt to get the free space of
# restricted mountpoints
# Privileged checks
# Required because check_disk may attempt to get the free space of
# restricted mountpoints
- name: Disk Usage
command: check_by_ssh!/usr/bin/sudo /usr/lib/nagios/plugins/check_disk -M -u GB -X nfs -X tracefs -X cgroup -X tmpfs -X overlay -X shm -w 15% -c 10% -W 15% -K 10% -A -I '^/run/' -I '^udev$' -I '^/var/lib/kubelet/' -I'^/tmp/.mount_'
# Device type checks
# R720
- name: CPU0 Temperature
command: check_by_ssh!/usr/bin/sudo /usr/local/bin/monitoring-scripts/check_temp -w 65 -c 75 --sensor coretemp-isa-0000
- name: CPU1 Temperature
command: check_by_ssh!/usr/bin/sudo /usr/local/bin/monitoring-scripts/check_temp -w 65 -c 75 --sensor coretemp-isa-0001
# Device role checks
# hypervisor (which is assumed to be Proxmox)
- name: PVE Unit pve-firewall.service

View File

@ -31,6 +31,7 @@
line: "nagios-checker ALL = (root) NOPASSWD: {{ item }}"
with_items:
- /usr/lib/nagios/plugins/check_disk
- /usr/local/bin/monitoring-scripts/check_temp
tags: [ nagios, sudo ]
- hosts: all
tasks:

@ -0,0 +1 @@
Subproject commit 1a332f6788d4ae24b52948850965358790861432