LibreNMS平均CPU使用率監控功能

  最近有一個需求是要去CPU的監控,但使用的方式為 

processors.processor_usage >= processor_perc_warn

這個方式去監控是依據每一顆CPU去做監控的.如果該主機有 8 顆CPU,

其中有 1 顆 超過告警值 即會發出告警,但其它 7 顆可能沒那麼重的 loading .

這種方式的監控似乎不是那麼合理.理想的情境應是依 這 8 顆的平均值去做告警才是.

在 官方 https://docs.librenms.org/Alerting/Rules/ 的 Advanced 處,

有看到一個 SQL 語法.

SELECT *,AVG(processors.processor_usage) as cpu_avg FROM devices,processors WHERE (devices.device_id = ? AND devices.device_id = processors.device_id) AND (devices.status = 1 && (devices.disabled = 0 && devices.ignore = 0)) = 1 HAVING AVG(processors.processor_usage) > 10

後來試一下真的可以用.

壓測方式

[root@librenms CPULoadGenerator]# python3.6 -m cpu_load_generator -l 0.8 -d 600 -c -1

使用工具

https://pypi.org/project/cpu-load-generator/

Project description

CPU Load Generator

This package allows to generate a fixed CPU load for a finite time period. 

The script takes in input the desired CPU load, the duration of the experiment 

and the CPU core or all cores on which the load has to be generated.

Python versions This master branch refers to Python versions 3.x

Dependencies Installment of psutil is required 

(it is already set as dependency in the setup.py)

Install psutil:

pip install psutil

Other dependencies for unit tests and linters:

pip install mock pytest flake8 tox

Usage To generate 20% of load on core 0 for 20 seconds run:

python -m cpu_load_generator -l 0.2 -d 20 -c 0

To generate 50% of load on all logical cores for 20 seconds run:

python -m cpu_load_generator -l 0.5 -d 20 -c -1

There is an option to run CPU load based on profile file. An exemplary profile 

is under load_profiles/default_profile.json. In order to run generator based on

a profile issue the following command:

python -m cpu_load_generator -p <path_to_profile_json>

To use the package features from python code:

Install the package from PyPi by issuing the following command:

python -m pip instal cpu-load-generator

To use its features from your code:

from cpu_load_generator import load_single_core, load_all_cores, from_profile

load_single_core(core_num=0, duration_s=20, target_load=0.4)  # generate load on single core (0)

load_all_cores(duration_s=30, target_load=0.2)  # generates load on all cores

from_profile(path_to_profile_json=r"c:\profiles\profile1.json")

DEMO