Performance Profiling for Python application running in Linux App Service

This article is contributed. See the original author and article here.

py-spy is a sampling profiler for Python programs. It lets you visualize what your Python program is spending time on without restarting the program or modifying the code in any way. py-spy is extremely low overhead: it is written in Rust for speed and doesn’t run in the same process as the profiled Python program. This means py-spy is safe to use against production Python code.

py-spy works on Linux, OSX, Windows and FreeBSD, and supports profiling all recent versions of the CPython interpreter (versions 2.3-2.7 and 3.3-3.9).

This article shows a demonstration of how to use py-spy to run sampling profiler for your Python application running in Linux App Service.

Step 1: Install py-spy for your Python application

In your Python application project, find your requirements.txt file. Add “py-spy” in the file.
This is the only change we need to make. No other code changes are required.

Step 2: Deploy your application to Azure Web App.

When deploying the code, Azure orxy build will help us install the py-spy package.

Step 3: Run live performance monitor for your application

Py-spy allow us attach to a running Python process, and monitor its performance.
The tool is similar to the Unix top command, it shows a live view of what functions are taking the most time in your python program, . Running py-spy with:

First use “ps aux | grep python” command to find your python process ID.

Run “py-spy top –pid <python process id>”

Here is my sample Python code. When I access http://<webapp-name>.azurewebsites.net/test2, I can see a lot cpu time consumed by my runjob2() function.

	import time
	from flask import Flask
	app = Flask(__name__)
	@app.route("/")
	def hello():
	    runjob()
	    return "Hello Azure, from Flask!"
	@app.route("/test1")
	def test1():
	    return "Hello Test1!"
	@app.route("/test2")
	def test2():
	    runjob2()
	    return "Hello Test2!"
	def runjob():
	    counter = []
	    for i in range(0,5):
	        test()
	        counter.append(i)
	        print (counter)
	def runjob2():
	    for i in range(0,10):
	        i = i+1
	        time.sleep(1)
	def test():
	    print(">>>>>start test")
	    time.sleep(1)
    print("<<<<<<<done")

Step 3: Use record command to record profiles to a flame graph file

Run the following command in your WEBSSH
root@a029b609b0ab:~# py-spy record -o /home/profile.svg –pid 38

It will start profiling record:

Run tests in your web browser
For example, access different web pages, http://<webapp-name>.azurewebsites.net/test1 and http://<webapp-name>.azurewebsites.net/test2

Use Control-C to exit the performance profiling.
The recorded svg file will be found in the path you defined in the “py-spy record” command

Go to your App Service File manager to download the .svg file
https://<webapp-name>.scm.azurewebsites.net/newui/fileManager#

Open the downloaded .svg file in any of web browser in your local machine. You will see the flame graph of your python application.
Flame graphs are a visualization of profiled software, allowing the most frequent code-paths to be identified quickly and accurately.