Pod's median CPU usage

Question

In order to adjust my pods limits and requests, I want to know their median CPU usage. But I can't understand how to get this value.

One of my queries is:

max_over_time(
    avg by (my_label) (
      rate(container_cpu_usage_seconds_total{labels}[3m])
        * on (pod) group_left (my_label) 
       kube_pod_labels{labels}
    )
[1d:])

Result is

How can I get only one median value, not the array? In this example, it will be something like ~0.9

Any help is appreciated.

Update.

After some research, I've made this query:

quantile_over_time(0.99,
    avg by (my_label) (
        rate(container_cpu_usage_seconds_total{labels,cpu="total"}[3m])
            * on (pod) group_left (my_label) 
        kube_pod_labels{labels})
[30d:15m])

Docs say:

The 0.5-quantile is known as the median. The 0.95-quantile is the 95th percentile.

For my purpose median isn't suitable, thus I decided to use 0.99-quantile.

This solves my problem, but if anyone can add something or comment my decision it would be great.

Pod's median CPU usage

0 Answers