From 8cb292a4b2bc7668229493c066965d73e2be89b6 Mon Sep 17 00:00:00 2001
From: "Dustin C. Hatch" <dustin@hatch.name>
Date: Thu, 11 Jul 2024 22:07:27 -0500
Subject: [PATCH] v-m: alerts: Add alert for temperatures

After the incident this week with the CPU overheating on _vmhost1_, I
want to make sure I know as soon as possible when anything is starting
to get too hot.
---
 victoria-metrics/alerts.yml | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/victoria-metrics/alerts.yml b/victoria-metrics/alerts.yml
index c51fe40..ce716a1 100644
--- a/victoria-metrics/alerts.yml
+++ b/victoria-metrics/alerts.yml
@@ -141,3 +141,10 @@ groups:
       - ignoring (instance) group_right (scope) (patroni_xlog_replayed_location != 0)
       > 10240
     for: 10m
+
+- name: Temperature
+  rules:
+  - alert: High Temperature
+    expr: >-
+      {__name__=~"collectd_.*_temperature", sensors!~"i350bb.*"} > 80
+    for: 10m