Disasters happen: coffee spills on a laptop, electronics are dropped, hard disks fail or become corrupt. The combination of Conda environments and version control provides a near-bullet-proof disaster mitigation strategy. We will also discuss disaster recovery for when disasters happen.
Topics covered in this post:
- Disaster Mitigation
- Disaster Recovery
Click here for a list of related posts.
Updated: 2020-05-03
Disaster Mitigation/Recovery
Having a method to regenerate a project’s Conda environment is very important. Instructions to generate an environment specifications file, which is used to recreate a Conda environment, are provided in the Conda Environments post. The specifications text file provides a quick way to recover a working Python environment after total data loss. Ensure that each project has an updated specifications text file within each project top-level directory.
You can also use a YAML environment file if exact reproducibility is not a concern. While YAML files are better for cross platform environments, you probably won’t be recovering into a different operating system than the one you used before disaster. I recommend having both a YAML environment file and an environment specifications file. That way you’re covered either way.
A solid disaster mitigation strategy is to have at least 3 working copies of your projects. One will be on the local computer, another copy will be on a remote Git repository (e.g. GitLab), and another should exist on either a cloud server (such as DropBox), or an external USB/hard drive (or use both, a cloud service and USB/hard drive). This strategy provides a local and remote backup in case your place of work/home burns down, or succumbs to some other unforeseen disaster.
Let’s say you’re having a bad day and your hard drive fails. All work is lost. If you had been pushing local Git commits to a remote repository, backing up the repository on another storage device (just in case the remote Git repository also fails), and had a Conda specifications file, then recovery is quick and easy (after a new hard drive and operating system are installed).
To recover, install Git on the restored local computer, create a Projects folder (e.g. ~/Documents/Projects/), and clone your remote Git repository into the Projects folder. Then, install Miniconda3 and recreate the Conda environment. Viola, you’re back in business.
Related Posts
- Python for Scientists – Install for Windows
- Python for Scientists – Install for Linux
- Python for Scientists – Environments
- Python for Scientists – Project Organization
