flask-pypprof

Blueprint for pprof profiling endpoints a la GO's that can be added to python Flask applications. flask-pypprof is a wrapper of pypprof and based on django-pypprof ideas.

Installation

pip install flask-pypprof

Usage

To add the pprof endpoints to your application, simply register the blueprint:

from flask_pypprof import get_pprof_blueprint

app.register_blueprint(get_pprof_blueprint())

Once registered, you will be able to access the following endpoints:

• /debug/pprof/profile: will return a CPU profile
• /debug/pprof/heap: will return a heap profile
• /debug/pprof/thread: will return a thread profile
• /debug/pprof/wall: will return a wall time profile(work in progress)

Configuration

You can configure the memory sample rate by setting the following environment variable:

• MEMORY_PROFILER_ENABLED: enables the memory profiler (default: false)
• MEMORY_SAMPLE_RATE: sets the memory profiling sample rate (default: 128 * 1024)

export MEMORY_PROFILER_ENABLED = false
export MEMORY_SAMPLE_RATE = 128 * 1024

Fetching profiles from your application

Fetch a 30 seconds CPU profile:

go tool pprof -http=:8088 http://localhost:8081/debug/pprof/profile?seconds=30

Fetch a heap profile:

go tool pprof -http=:8088 http://localhost:8081/debug/pprof/heap

Compatibility

Python 3.8, 3.9, 3.10 and Flask >= 2.0.0 and Linux only.

Known issues

• zprofile issue with python 3.11: https://github.com/timpalpant/zprofile/pull/2
• mprofile only works on linux/amd64 or at least linux/arm64 is not supported. Also see timpalpant/mprofile#3.

License

This code base is available under the Apache License, version 2.