How-tos

How to pack serverless Python actions

Share this post:

For my tutorial on automated data retrieval and analytics, I use IBM Cloud Functions to automatically fetch GitHub traffic statistics once a day. It is implemented as a serverless Python action. Because some Python packages are needed, the question was how to pack and create the action. In this blog post, I share my experiences.

Serverless access to Db2 and GitHub

Serverless access to Db2 and GitHub

Overview

IBM Cloud Functions offers three options to provide the code when creating a Python action.

  1. Upload a single file with the entire code.
  2. Provide a zip archive that has the source code and required modules.
  3. Create a zip archive with the virtualenv of the code and the environment with installed modules.
  4. An entirely different approach would be to build a Docker image with the Python code.

Option 1 typically works for smaller functions or projects that only use the built-in libraries. Depending on the chosen base environment, there is already a longer, useful list of Python libraries provided. Option 2 seems doable for few missing libraries and option 3 would be the last resort.

Packaging Python actions

Naturally, I started by looking into the easiest option first. The newer Jessie-based Python 3 environment features the IBM Db2 driver, support for Cloud Object Storage, Cloudant, Watson services, data science and analytics, and cryptography. However, it was missing the module I planned to use for simplified access to the GitHub API.

Thus, I started to compare creating a zip archive with either the required files or the virtual environment. The virtualenv is easy to create and to automate (with some experience), but I wanted to optimize the archive size. With only few files needed for my function, I downloaded the required libarary files and put them in a directory with my own code. Thereafter, I created the zip archive and was able to deploy the Python action.

Conclusions

If you write cloud functions that interact with IBM Cloud services, often no extra libaries are needed. They are available as part of the latest runtime environments. If a function is dependent on additional modules, downloading and packing them into a zip archive usually is straightforward and a good choice. For more complex dependencies, using a virtual environment provides the necessary support to succeed with Python actions.

Take a look at the tutorial and try out the serverless Python action. All the necessary code, including the zip archive with the packaged Python action, is available on GitHub.

If you have feedback, suggestions, or questions about this post, please reach out to me on Twitter (@data_henrik) or LinkedIn.

Technical Offering Manager / Developer Advocate

More How-tos stories
January 18, 2019

IBM Cloud Kubernetes Service Supports containerd

For IBM Cloud Kubernetes Service clusters that run version 1.11 or later, containerd replaces Docker as the container runtime for Kubernetes.

Continue reading

January 18, 2019

Strategies for Building Multi-Region Apps

Enterprises seek to minimize downtime and create resilient architectures to achieve maximum availability. In order to help achieve those goals, we added a new tutorial to our collection of solution tutorials that highlights IBM Cloud's capabilities for building resilient solutions.

Continue reading

January 11, 2019

Deploying to IBM Cloud Environments with IBM Cloud Developer Tools CLI

Learn how to deploy applications to IBM Cloud and IBM Cloud Private environments using IBM Cloud Developer Tools CLI version 2.1.12.

Continue reading