Statistics with python pandas

[LenR]

I'm not sure how useful this might be, but I was playing around after someone asked what would happen across the board if our CPU usage increased by 25%. The math isn't the point, the use of pandas is.

I use the pyzabbix api, so added the standard includes

import pandas as pd
import numpy as np

Then for each host in a prior selected host group:

- find the CPU Busy item (if it exists
- get a weeks trends
- The data returned by the API can be loaded directly into a pandas dataframe :-O
- calculate the avg, max and standard deviation
- calculate a worse case, where the max plus the std-dev increase by 25%
- then lest hosts that are currently running over 50% CPU Busy with the calculated scare values

for host in hosts:
cpu_item = zapi.item.get(hostids=host['hostid'],search={'name': 'CPU Busy'})
if len(cpu_item) == 0:
# print("No item for", host['name'])
continue
cpu_trend = zapi.trend.get(itemids=cpu_item[0]['itemid'],limit=168,output=["clock", "value_avg", "value_max"])
if len(cpu_trend) == 0:
# print("No trend for", host['name'])
continue
df = pd.DataFrame(cpu_trend)
cpu_avg_max = df['value_max'].astype(float).mean()
cpu_max_avg = df['value_avg'].astype(float).max()
cpu_avg_std = df['value_avg'].astype(float).std()
cpu_up_var = (cpu_max_avg + cpu_avg_std) * 1.25

if cpu_avg_max > 50:
print(host['name']+','+str(cpu_max_avg)+','+str(cpu_avg_max)+','+str (cpu_avg_std)+','+str(cpu_up_var))

[LenR]

Another example, using the same data, but list hosts that had more than 5 hours above 80% CPU busy for the past week:
'
df = pd.DataFrame(cpu_trend)
cpu_avg_max = df['value_max'].astype(float).mean()
freq = df['clock'][df['value_max'].astype(float) > 80].count()
if freq > 5:
print(host['name']+','+str(freq)+','+str(cpu_avg_max))