From Wikitech
Jump to navigation Jump to search

Several Python runtimes are available for use inside Toolforge:

For guidance on writing tools in Python, see e. g. Help:Toolforge/My first Flask OAuth tool and Help:Toolforge/My first Django OAuth tool.

Deprecating Python 2

Many legacy tools and code examples use Python 2.x as a runtime. Use of Python 3.x is encouraged for new code as Python 2.7 stopped being maintained in 2020. Toolforge will provide some amount of support for Python 2.x through 2022 because Debian will be supporting Python 2.7 in Debian 10 (buster). This support will only extend to critical security patches however.

Virtual environments

If you don't want to create one tool for each Python task you need specific environment for, that's when virtual environment comes handy. The advantages of virtual environments don't end there. Apart from packages you can also maintain several separated git/svn repositories, several command line profiles, etc. You basically create this small virtual unit in your folder system, set up its environment to your needs and then reach its contents when you need it.

The python3 venv must be created in the same execution environment that the venv will be used from. This means:

  • if the tool uses a a kubernetes backend (recommended), the venv should be boostrapped inside a container.
  • if the tool uses a grid engine backend, the venv should be bootstrapped directly on a Toolforge bastion filesystem like or

Read below for more concrete information.

For Kubernetes backend

This is for tools that use the Toolforge Kubernetes backend (recommended).

Kubernetes python webservices

See Toolforge webservices with python.

Kubernetes python jobs

Follow these instructions if you are using the Toolforge Jobs framework.

You need to bootstrap your python venv from inside a job itself (similar to what happens in kubernetes webservices).

Create a script similar to this:

# use bash strict mode
set -euo pipefail

# delete the venv, if it already exists
rm -rf pyvenv

# create the venv
python3 -m venv pyvenv

# activate it
source pyvenv/bin/activate

# upgrade pip inside the venv and add support for the wheel package format
pip install -U pip wheel

# Change the following section depending on what your tool needs!

# install some concrete packages
# pip install requests
# pip install pyyaml

# or, install all packages from src/requirements.txt
# pip install -r src/requirements.txt

Then run it in the desired python container, selecting the python version you prefer, example:

tools.mytool@tools-sgebastion-11:~$ ls
tools.mytool@tools-sgebastion-11:~$ chmod ug+x
tools.mytool@tools-sgebastion-11:~$ toolforge-jobs run bootstrap-venv --command "cd $PWD && ./" --image python3.9 --wait
tools.mytool@tools-sgebastion-11:~$ ls pyvenv

Now you can run your python tool using this venv, example:

tools.mytool@tools-sgebastion-11:~$ cat src/
import requests
r = requests.get('')
tools.mytool@tools-sgebastion-11:~$ toolforge-jobs run mytool --command "pyvenv/bin/python src/" --image python3.9
tools.mytool@tools-sgebastion-11:~$ cat mytool.out

For Grid Engine backend

This is retained as a reference for existing tools that use the Toolforge Grid Engine backend. Do not create new tools using these instructions.

Grid Engine python webservice

See Toolforge webservices with python.

Grid Engine python jobs

Follow these instructions if you are using the Grid Engine backend to run your jobs. Do not create new jobs using these instructions.

You can create your first virtual environment using:

$ python3 -mvenv my_venv

This will install package manager, some basic tools, commands and prerequisites, everything into your new little unit. Once you created one, let's use it and play with it:

$ source my_venv/bin/activate

(my_venv) $ pip install my_dream_package==7.0.3

Once you are happy with it, you can always leave using:

(my_venv) $ deactivate

This way you can create as many separated Python environments as you wish.

You can reach it again from the inside the same way any time you want. This is handy when you want to update it for example. But for scheduled tasks, you would have to create a batch file with multiple commands to reach it, use it and leave it.

Use venv with scheduled tasks

Since it is saved in a folder in your Toolforge space, you can always use it from the outside just like any other folder. Well, you can not alter it this way, but for scheduled tasks you usually don't need to:

$ jsub -N my_task -once -quiet my_venv/bin/python3 my_script
Use your venv everywhere

You can also use your virtual environment everywhere by default. You can activate it using .profile like:

$ echo "source my_venv/bin/activate" >> .profile

See also

Communication and support

Support and administration of the WMCS resources is provided by the Wikimedia Foundation Cloud Services team and Wikimedia Movement volunteers. Please reach out with questions and join the conversation:

Discuss and receive general support
Receive mail announcements about critical changes
Subscribe to the cloud-announce@ mailing list (all messages are also mirrored to the cloud@ list)
Track work tasks and report bugs
Use a subproject of the #Cloud-Services Phabricator project to track confirmed bug reports and feature requests about the Cloud Services infrastructure itself
Learn about major near-term plans
Read the News wiki page
Read news and stories about Wikimedia Cloud Services
Read the Cloud Services Blog (for the broader Wikimedia movement, see the Wikimedia Technical Blog)