Python SDK Instrumentation

This page will dive into the nitty gritty details on installing Rookout under various configurations.
If you are encountering any difficulties with deploying Rookout, this is the place to look.

Python

The Python SDK provides the ability to fetch debug data from a running application in real time.
It can easily be installed by running the following command:

pip install rook

Setup

Import the SDK within your application:

import rook

if __name__ == "__main__":
    rook.start(token='[Your Rookout Token]',
               labels={"env":"dev"}) # Optional,see Labels page below Projects
    # Your program starts here :)

The SDK should be imported just before the application begins executing.
This is due to the fact that in Python, there's no clean way to identify a module has finished defining it's classes.

For Pre-forking servers please read the relevant section..

SDK API

start

start(token=None,
    labels=None,
    host=None,
    port=None,
    debug=None,
    throw_errors=None,    
    log_to_stderr=None,
    git_commit=None,
    git_origin=None,
    **kwargs)

The start method is used to initialize the SDK in the background and accepts the following arguments:

Argument	Environment Variable	Default Value	Description
`token`	`ROOKOUT_TOKEN`	None	The Rookout token for your organization. Should be left empty if you are using a Rookout ETL Controller
`labels`	`ROOKOUT_LABELS`	{}	A dictionary of key:value labels for your application instances. Use `k:v,k:v` format for environment variables
`git_commit`	`ROOKOUT_COMMIT`	None	String that indicates your git commit or a branch name
`git_origin`	`ROOKOUT_REMOTE_ORIGIN`	None	String that indicates your git remote origin
`fork`	`ROOKOUT_ENABLE_FORK`	False	Set to `True` to enable support in forked processes
`host`	`ROOKOUT_CONTROLLER_HOST`	None	If you are using a Rookout ETL Controller, this is the hostname for it
`port`	`ROOKOUT_CONTROLLER_PORT`	None	If you are using a Rookout ETL Controller, this is the port for it
`proxy`	`ROOKOUT_PROXY`	None	URL to proxy server
`debug`	`ROOKOUT_DEBUG`	False	Set to `True` to increase log level to debug
`throw_errors`	None	False	Set to `True` to throw an exception if `start` fails (error message will not be printed in console)

restart

restart(labels=None)

The restart method is used to change the SDK labels:

Argument	Environment Variable	Default Value	Description
`labels`	`ROOKOUT_LABELS`	{}	A dictionary of key:value labels for your application instances. Use `k:v,k:v` format for environment variables

flush

flush()

The flush method allows explicitly flushing the Rookout logs and messages.

Test connectivity

To make sure the SDK was properly installed in your Python (virtual) environment, and test your configuration (environment variables only), run the following command:

python -m rook

Source Commit Detection

The Python SDK supports detecting the existing source code commit in the following methods, in descending order of priority:

If the environment variable “ROOKOUT_COMMIT” exists, use it.
If the environment variable “ROOKOUT_GIT” exists, search for the configuration of the “.git” folder and use its head.
If the main application is running from within a Git repository, use its head.

Supported Python versions

Implementation	Versions
CPython	2.7, 3.5, 3.6, 3.7, 3.8
PyPy	6.0.0

Rookout was tested on pip versions 9+.

Note: We recommend avoiding production deployments of Rookout on Windows OS.

Dependencies

The Python SDK contains native extensions. For most common interpreter and OS configurations, pre-built binaries are provided. For other configurations, a build environment is needed to successfully install Rookout.

If you encounter an error similar to the following example, be sure to install the environment specific build tools specified below:

    Could not find <Python.h>. This could mean the following:
      * You're on Ubuntu and haven't run `apt-get install python-dev`.
      * You're on RHEL/Fedora and haven't run `yum install python-devel` or
        `dnf install python-devel` (make sure you also have redhat-rpm-config
        installed)
      * You're on Mac OS X and the usual Python framework was somehow corrupted
        (check your environment variables or try re-installing?)
      * You're on Windows and your Python installation was somehow corrupted
        (check your environment variables or try re-installing?)

Here are the commands for installing the build environments for some common OS:

OS X

Debian

Fedora

Alpine

xcode-select --install
# If installing for PyPy on macOS, installing pkg-config is also required:
brew install pkg-config

apt-get update -q && apt-get install -qy g++ python-dev

yum install -qy gcc-c++ python-devel

apk update && apk add g++ python-dev

Pre-forking servers

Several popular application servers for Python load the application code during startup and then fork() the process multiple times to worker processes.

If you are using one of those servers, You can set the fork argument in the SDK api to true to automatically enable Rookout in forked processes, and no additional changes will be required.

If you are don't enable fork support Rookout must be started in each of the workers processes.
We have included sample snippets for a few common options:

uWSGI

Gunicorn

Celery

try:
    from uwsgidecorators import postfork

    # Run Rookout after the fork
    @postfork
    def run_rookout():
        import rook
        rook.start(token='[Your Rookout Token]')
except ImportError:
    # If there's no uWSGI, run Rookout normally
    import rook
    rook.start(token='[Your Rookout Token]')

You must also enable threads by adding --enable-threads to the command line or enable-threads = true in the uWSGI ini file.
Read more about it here.

# Gunicorn does not preload applications by default
# Under some configurations (such as --preload) you will need to create gunicorn_config.py file.

# Load the file using the -c flag: 'gunicorn -c python:gunicorn_config server:app'

def post_fork(server, worker):
    import rook
    rook.start(token='[Your Rookout Token]')

from celery.signals import worker_process_init

# Use the `worker_process_init` signal to load Rookout on worker start:
@worker_process_init.connect
def start_rook(*args, **kwargs):
    import rook
    rook.start(token='[Your Rookout Token]')

Serverless and PaaS deployments

Integrating with Serverless

When integrating Rookout into a Serverless application, you should explicitly flush the collected information.
For most common Serverless runtimes, Rookout provides easy to use wrappers such as:

from rook.serverless import serverless_rook

@serverless_rook
def lambda_handler(event, context):
  return "Hello world"

Note: Adding the Rookout SDK will slow down your Serverless cold-start times. Please make sure your timeout is no less then 10 seconds.

For more information, please check out our deployment-examples.

Building

If you are running your application on a Serverless or PaaS (Platform as a Service), you must build your package in an environment similar to those used in production. If you are running on a Windows or Mac machine (or using an incompatible Linux distribution) you may encounter some issues here.

Many Serverless frameworks (such as AWS SAM) have built-in support for it and will work out of the box.

If you need to set up your own build, we recommend using Docker, with a command line such as:

docker run -v `pwd`:`pwd` -w `pwd` -i -t lambci/lambda:build-python2.7 pip install -r requirements.txt

For more information check out this blog post: https://www.rookout.com/3_min_hack_for_building_local_native_extensions/

Python Spark (PySpark) applications

Import the SDK as usual in the main function that runs on the Spark driver.
To import the SDK on Spark executors, run spark-submit with --conf spark.python.daemon.module=rook.pyspark_daemon.
If running under YARN, specify the ROOKOUT_TOKEN environment variable for your application master and executor nodes like so:

spark-submit --conf spark.python.daemon.module=rook.pyspark_daemon --conf spark.yarn.appMasterEnv.ROOKOUT_TOKEN=[Your Rookout Token] --conf spark.executorEnv.ROOKOUT_TOKEN=[Your Rookout Token]

For more information, please check out our deployment-examples.