[PYTHON] How to display the CPU usage, pod name, and IP address of a pod created with Kubernetes

Introduction

Data acquisition is an important task in evaluating experiments using pods. In this article, I will write a program to get the pod name, IP address, and CPU usage.

Creating a pod

Duplicate a pod using kubernetes Deployment reprica.yaml sets the number of replicas to 4. A container is a program that displays a web application with an nginx image.

reprica.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: replicas-deployment
spec:
  replicas: 4
  selector:
    matchLabels:
      app: server-app
  template:
    metadata:
      labels:
        app: server-app
    spec:
      containers:
        - name: nginx-container
          image: nogisora/serverimage
          ports:
            - containerPort: 80
~                              

The created pod could be displayed like this.

kubectl get pod
NAME                                   READY   STATUS    RESTARTS   AGE
replicas-deployment-5b669df64c-wnfbp   1/1     Running   2          34d
replicas-deployment-5b669df64c-csscv   1/1     Running   2          34d
replicas-deployment-5b669df64c-fj6kr   1/1     Running   2          34d
replicas-deployment-5b669df64c-l8k7x   1/1     Running   2          34d

Next, write a program to get the data of this pod. api_cpu.py makes it possible to get the CPU usage rate, IP address, and pod name. In addition, the pod with the lowest CPU usage is displayed below.

• The json part is commented out because it was not necessary in the experiment.

api_cpu.py

from kubernetes import client, config
import json

config.load_kube_config()
api = client.CustomObjectsApi()
v1 = client.CoreV1Api()
resource = api.list_namespaced_custom_object(group="metrics.k8s.io",version="v1beta1", namespace="c0118220", plural="pods")

all_array = []

#print(json.dumps(resource,ensure_ascii=False, indent=4, sort_keys=True, separators=(',', ': ')))
for pod in resource["items"]:
    s=pod["metadata"]["name"],pod['containers'][0]["usage"]["cpu"].replace('n','')
    all_array.append(s)
    print(s)

max=None
max_name=""
for container in all_array:
    if max is None:
        max = int(container[1])
    elif int(container[1])<max:
        max=int(container[1])
        max_name=container[0]
print(max_name,max)

# Configs can be set in Configuration class directly or using helper utility
print("Listing pods with their IPs:")
ret = v1.list_pod_for_all_namespaces(watch=False)
for i in ret.items:
    if 'replicas-deployment' in i.metadata.name:

        print("%s\t%s\t%s" % (i.status.pod_ip, i.metadata.namespace, i.metadata.name))
~                                                                                ~                                                            

The execution result is as follows. The load is applied by the load experiment tool Locust. The acquired data can be displayed.

('replicas-deployment-5b669df64c-fj6kr', '28540182')
('replicas-deployment-5b669df64c-l8k7x', '30169433')
('replicas-deployment-5b669df64c-wnfbp', '31286353')
('replicas-deployment-5b669df64c-csscv', '27848575')
replicas-deployment-5b669df64c-csscv 27848575
Listing pods with their IPs:
10.42.1.112     c0118220        replicas-deployment-5b669df64c-wnfbp
10.42.1.113     c0118220        replicas-deployment-5b669df64c-csscv
10.42.2.191     c0118220        replicas-deployment-5b669df64c-fj6kr
10.42.2.198     c0118220        replicas-deployment-5b669df64c-l8k7x

This is the end of this article. In the future, I would like to add a program that can display only the pod with the lowest CPU usage.