Shulou(Shulou.com)06/01 Report--

In this issue, the editor will bring you about how to use the Kubernetes application performance analysis tool Kubectl Flame. The article is rich in content and analyzes and describes for you from a professional point of view. I hope you can get something after reading this article.

What is performance analysis?

Performance analysis is a common method to analyze application performance to improve code quality. The most popular visual performance analysis method is to generate flame diagrams.

Flame Diagram of Java Application based on Spring Framework

The y-axis is the stack depth, and the x-axis spans the sample population. Each rectangle is a function in which the width shows how often it appears, and sorting from left to right is not important (the stack is sorted alphabetically).

Performance Analysis on Kubernetes

Performance analysis is a complex task, and most profilers have two main problems:

The application needs to be modified, usually by adding flags to the execution command or importing some performance analysis libraries into the code.

Because performance is seriously affected during the analysis process, performance analysis in a production environment is usually avoided.

Choosing the right profiler may solve these problems, but this requires careful study and usually depends on the programming language and operating system.

It becomes more difficult to perform analysis on applications running in a Kubernetes cluster. You need to deploy a new container image that contains changes to the configuration file, rather than the currently running container. In addition, some performance problems may disappear when the application is restarted, which makes debugging difficult.

Kubectl flame

Kubectl Flame is a kubectl plug-in that can generate flame diagrams with low overhead. To analyze application performance without any application modification or downtime.

Project warehouse address: https://github.com/VerizonMedia/kubectl-flame

Installation

You can install the kubectl flame plug-in through Krew, and once you have installed Krew, you can install it with the following command:

$kubectl krew install flame

Use requirements

Supported languages: Go, Java (any JVM-based language), Python, and Ruby

Kubernetes cluster using Docker as a container runtime (tested on GKE,EKS and AKS)

Operation principle

Kubectl-flame initiates performance analysis by launching a profiler on the same node as the target container, and most profilers will share some resources with the target container: for example, enabling PID namespace sharing by setting hostPID to true, and enabling file system sharing by mounting / var/lib/docker and querying overlayFS.

In the background, kubectl-flame uses async-profiler to generate flame diagrams for Java applications, interacts with the target JVM through a shared / tmp folder, Golang supports ebpf-based analysis, and Python supports py-spy-based analysis.

Overview of kubectl flame

Use an example

1. Analyze Kubernetes Pod

Analyze the Java application mypod for 1 minute, and save the flame image to / tmp/flamegraph.svg:

$kubectl flame mypod-t 1m-- lang java-f / tmp/flamegraph.svg

two。 Analyze alpine-based containers

You need to use the-- alpine flag to analyze Java applications in alpine-based containers:

$kubectl flame mypod-t 1m-f / tmp/flamegraph.svg-lang Java-alpine

Note: this-- alpine flag is required only for Java applications, but not for Go analysis.

3. Analyze sidecar Container

A Pod containing multiple containers needs to specify the target container as a parameter:

$kubectl flame mypod-t 1m-- lang go-f / tmp/flamegraph.svg mycontainer above is how the Kubernetes application performance analysis tool Kubectl Flame shared by the editor is used. If you happen to have similar doubts, you might as well refer to the above analysis to understand. If you want to know more about it, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.