Python program logs are not displayed in k8s cluster

When deploying Python applications using Docker and Kubernetes, you may experience inconsistent log output. Specifically, usedocker logsYou can see the log output normally, but it is used in Kuberneteskubectl logsBut there is no log output. This article will analyze this problem in detail and provide a solution.

Problem description

Suppose we have a simple Python application, the contents are as follows:

import time
import logging

(level=)
logger = (__name__)

def main():
    while True:
        ("This is a test log message")
        (5)

if __name__ == "__main__":
    main()

Docker runs

Run the application using Docker:

docker build -t my-python-app .
docker run -d --name my-python-container my-python-app
docker logs my-python-container

at this time,docker logs my-python-containerThe logs can be output normally:

INFO:__main__:This is a test log message
INFO:__main__:This is a test log message
...

Kubernetes Deployment

Deploy the application to a Kubernetes cluster:

apiVersion: v1
kind: Pod
metadata:
  name: my-python-app
spec:
  containers:
  - name: my-python-container
    image: my-python-app
    command: ["python", ""]

Application configuration file:

kubectl apply -f 
kubectl logs my-python-app

at this time,kubectl logs my-python-appBut no logs were output.

Problem analysis

Standard output buffering

Python's standard output (stdout) and standard error stream (stderr) are buffered by default. This means that when Python writes to the log, it first stores the log data in the memory buffer instead of outputting it to the terminal or log system immediately. When the buffer is full or the program exits, the data in the buffer will be flushed to the output stream.

In Docker, run directlydocker runWhen  , the standard output is usually refreshed to the terminal immediately, so the log output can be seen normally. But in Kubernetes,kubectl logsRely on the container's standard output and error streams, if these streams are buffered,kubectl logsThe log may not be obtained in time.

Solution

To solve this problem, you can use-uOption to start the Python interpreter and force disable the output buffering. The specific methods are as follows:

Modify Dockerfile

Use in Dockerfile-uOptions start Python:

FROM python:3.9-slim

WORKDIR /app
COPY  .

CMD ["python", "-u", ""]

Modify Kubernetes configuration

Or modify it in the Kubernetes configuration filecommandFields:

apiVersion: v1
kind: Pod
metadata:
  name: my-python-app
spec:
  containers:
  - name: my-python-container
    image: my-python-app
    command: ["python", "-u", ""]

verify

Rebuild the Docker image and deploy it to Kubernetes:

docker build -t my-python-app .
docker push my-python-app
kubectl apply -f 
kubectl logs my-python-app

at this time,kubectl logs my-python-appThe log should be output normally.

Summarize

The inconsistent log output problem of Python programs in Docker and Kubernetes is mainly caused by the standard output buffering mechanism. By using-uOption to start the Python interpreter, which can disable output buffering, ensure that the logs can be output to the standard output stream in time, and thus be used in Kuberneteskubectl logsCheck the log normally.

This is the article about the content of the python program not displaying logs in the k8s cluster. For more related content of the python k8s log, please search for my previous articles or continue browsing the related articles below. I hope everyone will support me in the future!