Austin: Python frame stack sampler for CPython

A Frame Stack Sampler for CPython

Synopsis

Austin is a Python frame stack sampler for CPython written in pure C. It samples the stack traces of a Python application so that they can be visualised and analysed. As such, it serves the basis for building powerful profilers for Python.

Key features are:

• no instrumentation required;
• minimal impact on the performance of the profiling target;
• fast and lightweight;
• time and memory profiling modes.

The most interesting use of Austin is probably in conjunction with FlameGraph to profile Python applications while they are running, without the need of instrumentation. This means that Austin can be used on production code with little or even no impact on performance.

However, the output format can be grabbed from any other external tool for further processing. Look, for instance, at the following Python TUI, similar in spirit to py-spy .

Installation

Austin is available from the major software repositories of the most popular platforms.

On Linux, it can be installed using autotools or as a snap from the Snap Store. The latter will automatically perform the steps of the autotools method with a single command. On distributions derived from Debian, Austin can be installed from the official repositores with Aptitude.

On Windows, Austin can be easily installed from the command line from the Chocolatey repositories.

For any other platform, compiling Austin from sources is as easy as cloning the repository and running the C compiler.

With autotools

Installing Austin using autotools amounts to the usual ./configure , make and make install finger gymnastic. The only dependency is the standard C library.

git clone --depth=1 https://github.com/P403n1x87/austin.git
autoreconf --install
./configure
make
make install

Compilation has been tested with GNU GCC 7.3.0 on Linux, MinGW 2.28-1 on Windows and LLVM 8.0.0 with clang-800.0.42.1. The code is so simple that it really compiles with just

gcc -O3 -Wall src/*.c -o src/austin

so you can use just this command if you don't have autoreconf installed.

Add -DDEBUG if you want a more verbose syslog output on UNIX-like systems, or %TEMP%/austin.log on Windows.

From the Snap Store

Austin can be installed from the Snap Store with the following command

sudo snap install austin --classic

On Debian and derivatives

On March 30 2019, Austin has been accepted into the official Debian repositories and can therefore be installed with apt . The first Ubuntu release to include Austin is 19.10.

From Chocolatey

To install Austin from Chocolatey , run the following command from the command line or from PowerShell

choco install austin

To upgrade run the following command from the command line or from PowerShell:

choco upgrade austin

From sources

To install Austin from sources using the GNU C compiler, without autotools , clone the repository with

git clone --depth=1 https://github.com/P403n1x87/austin.git

and then run

gcc -s -Wall -O3 -o src/austin src/*.c

An extra flag is required on Windows, so the command to use in this case is

gcc -s -Wall -O3 -o src/austin src/*.c -lpsapi

Usage

Usage: austin [OPTION...] command [ARG...]
Austin -- A frame stack sampler for Python.

  -a, --alt-format           Alternative collapsed stack sample format.
  -e, --exclude-empty        Do not output samples of threads with no frame
                             stacks.
  -f, --full                 Produce the full set of metrics (time +mem -mem).
  -i, --interval=n_us        Sampling interval (default is 500us).
  -m, --memory               Profile memory usage.
  -o, --output=FILE          Specify an output file for the collected samples.
  -p, --pid=PID              The the ID of the process to which Austin should
                             attach.
  -s, --sleepless            Suppress idle samples.
  -t, --timeout=n_ms         Approximate start up wait time. Increase on slow
                             machines (default is 100ms).
  -?, --help                 Give this help list
      --usage                Give a short usage message
  -V, --version              Print program version

Mandatory or optional arguments to long options are also mandatory or optional
for any corresponding short options.

The output is a sequence of frame stack samples, one on each line. The format is the collapsed one that is recognised by FlameGraph so that it can be piped to flamegraph.pl in order to produce flame graphs, or redirected to a file for some further processing.

Normal mode

By default, each line has the following structure:

Thread [tid];[func] ([mod]);#[line no];[func] ...;L[line no] [usec]

The reason why the line number is not included in the ([mod]) part, as done by py-spy, is that, this way, the flame graph will show the total time spent at each function, plus the finer detail of the time spent on each line. A drawback of this format is that frame stacks double in height. If you prefer something more conventional, you can use the -a option to switch to the alternative format

Thread [tid];[func] ([mod]:[line no]);#[line no];[func] ... ([mod]:[line no]) [usec]

Memory and Full modes

When profiling in memory mode with the -m or --memory switch, the metric value at the end of each line is the memory delta between samples, measured in KB. In full mode ( -f or --full switches), the last three values on each line are the time delta, any positive memory delta (memory allocations) or zero and any negative memory delta (memory releases) or zero, i.e.

Thread [tid];[func] ([mod]:[line no]);#[line no];[func] ... ([mod]:[line no]) [usec] [+KB] [-KB]

Logging

Austin uses syslog on Linux and Mac OS, and %TEMP%\austin.log on Windows for log messages so make sure to watch these to get execution details and statistics. Bad frames are output together with the other frames. In general, entries for bad frames will not be visible in a flame graph as all tests show error rates below 1% on average.

Compatibility

Austin has been tested on the following systems (both 32- and 64-bit, unless otherwise specified).

NOTEAustin might work with other versions of Python on all the platforms and architectures above. So it is worth giving it a try even if your system is not listed below.

Linux

• Python 2.3 (2.3.7) on Ubuntu 18.04.1

• Python 2.4 (2.4.6) on Ubuntu 18.04.1

• Python 2.5 (2.5.6) on Ubuntu 18.04.1

• Python 2.6 (2.6.9) on Ubuntu 18.04.1

• Python 2.7 (2.7.15rc1) on Ubuntu 18.04.1

• Python 3.3 (3.3.7) on Ubuntu 18.04.1

• Python 3.4 (3.4.9+) on Ubuntu 18.04.1

• Python 3.5 (3.5.2) on Ubuntu 18.04.1

• Python 3.6 (3.6.5, 3.6.6, 3.6.7) on Ubuntu 18.04.x

• Python 3.7 (3.7.0, 3.7.1, 3.7.3, 3.7.4) on Ubuntu 18.04.x

Mac OS

• Python 2.7 (2.7.10) on OS X "El Capitan" 10.11.4

• Python 3.7 (3.7.0) on OS X "El Capitan" 10.11.4

Due to the introduction of the System Integrity Protection, Austin cannot be used to profile Python applications that run using binaries located in system folders. The simplest solution is to create a virtual environment and use the local Python binaries instead.

Windows

• Python 2.7 (2.7.13) on windows 10 64-bit

• Python 3.6 (3.6.6) on Windows 10 64-bit

• Python 3.7 (3.7.0) on Windows 10 64-bit

• Python 3.6 (3.6.5, 3.6.6) on Ubuntu 18.04 x86-64 via WSL

Why Austin

When there already are similar tools out there, it's normal to wonder why one should be interested in yet another one. So here is a list of features that currently distinguish Austin.

• Written in pure CAustin is written in pure C code. There are no dependencies on third-party libraries with the exception of the standard C library and the API provided by the Operating System.

• Just a samplerAustin is just a frame stack sampler. It looks into a running Python application at regular intervals of time and dumps whatever frame stack it finds. The samples can then be analysed at a later time so that Austin can sample at rates higher than other non-C alternative that also analyse the samples as they run.

• Simple output, powerful toolsAustin uses the collapsed stack format of FlameGraph that is easy to parse. You can then go and build your own tool to analyse Austin's output. You could even make a player that replays the application execution in slow motion, so that you can see what has happened in temporal order.

• Small sizeAustin compiles to a single binary executable of just a bunch of KB.

• Easy to maintainOccasionally, the Python C API changes and Austin will need to be adjusted to new releases. However, given that Austin, like CPython, is written in C, implementing the new changes is rather straight-forward.

Examples

The following flame graph has been obtained with the command

./austin -i 50 ./test.py | ./flamegraph.pl --countname=us > test.svg

where the sample test.py script has the following content

import psutil

for i in range(1000):
  list(psutil.process_iter())

Austin TUI

The Python TUI that is currently included in this repository provides an example of how to use Austin to profile Python applications. You can use PageUp and PageDown to navigate the frame stack of each frame as the Python application runs.

If you want to give it a go you can install it using pip with

pip install git+https://github.com/P403n1x87/austin.git

and run it with

austin-tui [OPTION...] command [ARG...]

with the same command line as Austin.

The TUI is based on python-curses . The version included with the standard Windows installations of Python is broken so it won't work out of the box. A solution is to install the the wheel of the port to Windows from this page. Wheel files can be installed directly with pip , as described in the linked page.

Web Austin

Web Austin is yet another example of how to use Austin to make a profiling tool. It makes use of d3-flame-graph to display a live flame graph in the web browser that refreshes every 3 seconds with newly collected samples. Web Austin can also be used for remote profiling by setting the WEBAUSTIN_HOST and WEBAUSTIN_PORT environment variables.

If you want to give it a go you can install it using pip with

pip install git+https://github.com/P403n1x87/austin.git

and run it with

austin-web [OPTION...] command [ARG...]

with the same command line as Austin. This starts a simple HTTP server that serves on WEBAUSTIN_HOST if set or on localhost otherwise. The port can be controlled with the WEBAUSTIN_PORT environment variable. If it is not set, Web Austin will use an ephemeral port.

Contribute

If you like Austin and you find it useful, there are ways for you to contribute.

If you want to help with the development, then have a look at the open issues and have a look at the contributing guidelines before you open a pull request.

You can also contribute to the development of Austin by either becoming a Patron on Patreon

by buying me a coffee on BMC

or by chipping in a few pennies on PayPal.Me .

License

GNU GPLv3