rotki Documentation

Appearance

Code Profiling ​

Python ​

Flamegraph Profiling for Tests ​

To use the flamegraph profiler, follow these steps:

  1. Install the latest dependencies required for profiling:

    sh
    pip install -r requirements_profiling.txt

  2. Install the flamegraph package in your system. Some OSes, such as Archlinux, have ready-made packages.

To profile a test run, add --profiler=flamegraph-trace to the pytest arguments. Once the test concludes, this will add a data file under /tmp with the data generated by the run. For example: /tmp/20211127_1641_stack.data.

Then, you can run the flamegraph tool on that data to generate an SVG:

sh
flamegraph.pl --title "rotki-test graph" /tmp/20211127_1641_stack.data > profile.svg

Finally, open the SVG with any compatible viewer and explore the flamegraph. It will look like this:

A flamegraph profiling example

Viztracer ​

Viztracer is a good tool for profiling the actual code as it runs. You can install it with:

sh
pip install viztracer

Then, run rotki's dev mode and add profiling arguments for Viztracer:

sh
pnpm run dev --profiling-args "-m viztracer --min_duration 0.2ms"

This will produce a result.json file in the main directory. The --min_duration argument is needed to avoid capturing data every nanosecond, which could result in a very large JSON file. You may need to adjust the arguments based on your needs.

To open and study the file, you can use vizviewer. For example:

sh
vizviewer --flamegraph result.json

to get a flamegraph, or simply:

sh
vizviewer result.json

to get the normal view.

For more information, check the Viztracer documentation.

py-spy ​

py-spy is a tool that generates a flamegraph of a running Python process and can attach to it. You can get it from here.

You can run the developer version or a normal binary and find its PID with a command like:

sh
ps aux | grep rotki

Then, you can attach to it with a command like:

sh
sudo py-spy record -o py-spy.profile.svg --pid 60243

assuming the PID is 60243.

You can also run rotki's dev mode directly by specifying the profile command and its arguments like this:

sh
pnpm run dev --profiling-cmd py-spy --profiling-args "record -o py-spy-profile.svg --"

Once done, there will be an SVG output at py-spy.profile.svg, which you can view with any SVG viewer, including a browser, and study the flamegraph.