summaryrefslogtreecommitdiffstats
path: root/health/health.d/cpu.conf
diff options
context:
space:
mode:
authorEmmanuel Vasilakis <mrzammler@mm.st>2021-04-20 16:24:41 +0300
committerGitHub <noreply@github.com>2021-04-20 16:24:41 +0300
commitf5bd20e60ae8a65b9c709fcd28d08c1bea268f2d (patch)
tree8897d68a69f82d8fd0732aca8c344778a5e62340 /health/health.d/cpu.conf
parent0a6a14e323ee3b7a2dc17b2ca5b0cce4a8b2eb5e (diff)
Provide new attributes in health conf files (#10961)
* read and store new attributes (class, component, type) from health conf files. Replace family variable in info strings * provide the attributes to jsons * remove extra semicolon * populate conf files with new attributes * added newline * remove extra defines from health.h * remove empty line * remove realloc * use helper variables for find_and_replace. Adjust position for next strstr * remove comments * Add type to mysql.conf and vcsa.conf * fix formatting * add parenthesis * remove extra assignment * changes to mysql_galera_cluster_state from master * add type Errors to unbound_request_list_overwritten * fix identation for info strings spawning more than one line * check for null, replace with empty string if true * add class, component, type to systemdunits.conf
Diffstat (limited to 'health/health.d/cpu.conf')
-rw-r--r--health/health.d/cpu.conf108
1 files changed, 60 insertions, 48 deletions
diff --git a/health/health.d/cpu.conf b/health/health.d/cpu.conf
index 32c69f8f5b..d112157683 100644
--- a/health/health.d/cpu.conf
+++ b/health/health.d/cpu.conf
@@ -1,55 +1,67 @@
# you can disable an alarm notification by setting the 'to' line to: silent
-template: 10min_cpu_usage
- on: system.cpu
- os: linux
- hosts: *
- lookup: average -10m unaligned of user,system,softirq,irq,guest
- units: %
- every: 1m
- warn: $this > (($status >= $WARNING) ? (75) : (85))
- crit: $this > (($status == $CRITICAL) ? (85) : (95))
- delay: down 15m multiplier 1.5 max 1h
- info: average CPU utilization over the last 10 minutes (excluding iowait, nice and steal)
- to: sysadmin
+ template: 10min_cpu_usage
+ on: system.cpu
+ class: System
+component: CPU
+ type: Utilization
+ os: linux
+ hosts: *
+ lookup: average -10m unaligned of user,system,softirq,irq,guest
+ units: %
+ every: 1m
+ warn: $this > (($status >= $WARNING) ? (75) : (85))
+ crit: $this > (($status == $CRITICAL) ? (85) : (95))
+ delay: down 15m multiplier 1.5 max 1h
+ info: average CPU utilization over the last 10 minutes (excluding iowait, nice and steal)
+ to: sysadmin
-template: 10min_cpu_iowait
- on: system.cpu
- os: linux
- hosts: *
- lookup: average -10m unaligned of iowait
- units: %
- every: 1m
- warn: $this > (($status >= $WARNING) ? (20) : (40))
- crit: $this > (($status == $CRITICAL) ? (40) : (50))
- delay: down 15m multiplier 1.5 max 1h
- info: average CPU iowait time over the last 10 minutes
- to: sysadmin
+ template: 10min_cpu_iowait
+ on: system.cpu
+ class: System
+component: CPU
+ type: Utilization
+ os: linux
+ hosts: *
+ lookup: average -10m unaligned of iowait
+ units: %
+ every: 1m
+ warn: $this > (($status >= $WARNING) ? (20) : (40))
+ crit: $this > (($status == $CRITICAL) ? (40) : (50))
+ delay: down 15m multiplier 1.5 max 1h
+ info: average CPU iowait time over the last 10 minutes
+ to: sysadmin
-template: 20min_steal_cpu
- on: system.cpu
- os: linux
- hosts: *
- lookup: average -20m unaligned of steal
- units: %
- every: 5m
- warn: $this > (($status >= $WARNING) ? (5) : (10))
- crit: $this > (($status == $CRITICAL) ? (20) : (30))
- delay: down 1h multiplier 1.5 max 2h
- info: average CPU steal time over the last 20 minutes
- to: sysadmin
+ template: 20min_steal_cpu
+ on: system.cpu
+ class: System
+component: CPU
+ type: Latency
+ os: linux
+ hosts: *
+ lookup: average -20m unaligned of steal
+ units: %
+ every: 5m
+ warn: $this > (($status >= $WARNING) ? (5) : (10))
+ crit: $this > (($status == $CRITICAL) ? (20) : (30))
+ delay: down 1h multiplier 1.5 max 2h
+ info: average CPU steal time over the last 20 minutes
+ to: sysadmin
## FreeBSD
-template: 10min_cpu_usage
- on: system.cpu
- os: freebsd
- hosts: *
- lookup: average -10m unaligned of user,system,interrupt
- units: %
- every: 1m
- warn: $this > (($status >= $WARNING) ? (75) : (85))
- crit: $this > (($status == $CRITICAL) ? (85) : (95))
- delay: down 15m multiplier 1.5 max 1h
- info: average CPU utilization over the last 10 minutes (excluding nice)
- to: sysadmin
+ template: 10min_cpu_usage
+ on: system.cpu
+ class: System
+component: CPU
+ type: Utilization
+ os: freebsd
+ hosts: *
+ lookup: average -10m unaligned of user,system,interrupt
+ units: %
+ every: 1m
+ warn: $this > (($status >= $WARNING) ? (75) : (85))
+ crit: $this > (($status == $CRITICAL) ? (85) : (95))
+ delay: down 15m multiplier 1.5 max 1h
+ info: average CPU utilization over the last 10 minutes (excluding nice)
+ to: sysadmin