Question 1

py-spy vs cProfile: which is better for production debugging?

Accepted Answer

py-spy is superior for production because it runs externally with minimal overhead, while cProfile modifies code and adds significant performance impact. Use cProfile for detailed, development-time profiling where code changes are acceptable.

Question 2

How to profile a Python app running in Docker with py-spy?

Accepted Answer

You need to add the SYS_PTRACE capability to the Docker container. Use --cap-add SYS_PTRACE when running the container or edit the docker-compose file to include it in the security context. Without this, py-spy will fail with permission errors.

Question 3

Can py-spy profile Cython extensions on macOS?

Accepted Answer

Yes, but only if the extensions are compiled with symbols and the generated C file is available for line numbers. However, the README notes that native extension support on macOS is limited compared to Linux and Windows, so check the compatibility table.

Question 4

Why does py-spy sometimes require sudo on Linux?

Accepted Answer

On Linux, attaching to an existing process requires root permissions due to ptrace restrictions. You can avoid this by having py-spy spawn the process or by adjusting the ptrace_scope sysctl variable to allow non-root attachments.

Question 5

How to generate a flame graph from a running Python process with py-spy?

Accepted Answer

Use the command 'py-spy record -o profile.svg --pid <PID>' to capture samples and output an interactive SVG flame graph. You can adjust the sampling rate with --rate for more detailed or faster profiles.

Question 6

Does py-spy work with Python 3.12 or newer versions?

Accepted Answer

Yes, py-spy supports all recent CPython versions up to 3.13, as stated in the README. It uses generated Rust structures to handle ABI changes, ensuring compatibility across updates.

Question 7

How to avoid pausing the Python program during profiling?

Accepted Answer

Use the --nonblocking flag, but be aware that this may cause increased error rates or partial stack traces due to non-atomic memory reads, as mentioned in the FAQ section on this option.

py-spy

What is py-spy?

Overview

Use Cases

Best For

Related Projects

Found a gem we're missing?

Not Ideal For

Pros & Cons

Pros

Cons

Frequently Asked Questions