profiling.md

Profiling

cProfile

cProfile is a deterministic profiler in which trace functions are executed at various points of interest (e.g. function call, function return, exceptions), and precise timings of these events are recorded.

To profile the application using cProfile, run make profile. Profiling information will be collected per-request in the profiling directory where it can be examined using the pstats interactive browser.

To load the file into the interactive browser where it can be sorted and queried as required run:

python -m pstats [filename].prof

Useful commands:

# Strip the long file paths to make the reports easier on the eyes:
strip

# Sort on internal time:
sort time

# Sort on total time:
sort tottime

# Show the top 10 functions:
stats 10

# Show all the documented commands:
help

# Get more help on a command, e.g. the stats command:
help stats

Combining profiles

The profiles can also be combined to give an overview of the profile between all requests.

Combine all the profiles in the profiling directory using:

pipenv run python scripts/merge_profiles.py

This will create a file called combined_profile.prof

To visualise this profile, snakeviz or gprof2dot can be used.

Visualisation

There are many handy profilers, but they lack a nice visualisation interface. We use snakeviz and gprof2dot for this.

SnakeViz

SnakeViz is a browser-based graphical visualisation tool to display profiles using Icicle and Sunburst plots. It also includes IPython line and cell magics that can help profile a single line or code blocks directly and then visualise them.

Install using:

pip install snakeviz

Visualise a profile:

snakeviz profiling/[filename].prof

For example:

snakeviz profiling/GET.questionnaire.31ms.1571053121.prof

gprof2dot

Converts profiler data into call graphs. It allows to filter functions based on metrics threshold and colour them nicely with hotspots.

First install graphviz and gprof2dot:

brew install graphviz
pip install gprof2dot

Visualise a profile:

gprof2dot -f pstats profiling/[filename].prof | dot -Tpng -o output.png

For example:

gprof2dot -f pstats profiling/GET.questionnaire.31ms.1571053121.prof | dot -Tpng -o profile.png

To visualise the combined profile run:

gprof2dot -f pstats combined_profile.prof | dot -Tpng -o combined_profile.png

---

Py-Spy

Py-Spy is a sampling profiler where instead of tracking every event (e.g. function call), the application is periodically interrupted, and stack snapshots are collected. The function call stack is then analysed to deduce the execution time of different parts of the application.

Deterministic profilers modify application execution in some way: profiling code is typically run inside the target Python process, which often slows down application execution. To avoid this performance impact, Py-Spy doesn’t run in the same process as the profiled Python program. Because of this, Py-Spy can be used in a production environment with little overhead.

Py First, install Py-Spy:

pip install py-spy

Run survey runner using make run and get the process id of the child process. There will be 2 flask processes running, the child process will the have the higher process id.

Get process id using:

ps aux | grep flask

Run Py-Spy using:

sudo py-spy record -o profile.svg --pid 12345

This will sample the application until the process is terminated. Once terminated, it will display a flame graph for the sampled profiles.

Why do you need to run Py-Spy as sudo?

---

More info: Python profiling tools