Feature Request: Add backup, restore, and prune commands to 'tutor images'

[sagrodat]

Is your feature request related to a problem? Please describe.

When developing custom themes or MFEs locally (using dev mode) and iterating through them quickly (moving changes from dev mode on Machine1 to prod environment - local mode on Machine2) I find myself having to do a combination of actions repeatedly, pretty often. Namely:

• creating an additional (backup) tag for currently used images in prod environment in case the new pull breaks something
• restoring to the previous (backed up) version of the image if the new pulled image broke something (because of discrepancies between the dev and local environment).
• deleting (pruning) old image tags no longer required since newly pulled image works correctly

Describe the solution you'd like

A new subset of commands for the tutor images ... command.
backup

Usage: tutor images backup [OPTIONS] {openedx|openedx-dev|mfe|all}

  Create a backup tag for the current images.

Options:
  -t, --tag TEXT  Tag suffix (default: BACKUP)
  -h, --help      Show this message and exit.

restore

Usage: tutor images restore [OPTIONS] {openedx|openedx-dev|mfe|all}

  Restore a broken image from a backup tag.

Options:
  -t, --tag TEXT  Tag suffix to restore from
  -h, --help      Show this message and exit.

prune

Usage: tutor images prune [OPTIONS] {openedx|openedx-dev|mfe|all}

  Remove local backup images to free up space.

Options:
  -t, --tag TEXT  Tag suffix to remove
  -h, --help      Show this message and exit.

##Describe alternatives you've considered
Well, up until a point I was doing all of that by hand using

docker tag OLD_IMAGE_TAG NEW_IMAGE_TAG

Recently after being fed up with the amount of manual input it required I looked up the docs and implemented a tutor plugin as follows. If it's handy enough maybe this could become part of Tutor itself.

import click
import subprocess
from tutor.commands.images import images_command
from utils import get_tutor_config

# 1. Map simple CLI names to Tutor's internal configuration keys
IMAGE_KEYS = {
    "openedx": "DOCKER_IMAGE_OPENEDX",
    "openedx-dev": "DOCKER_IMAGE_OPENEDX_DEV",
    "mfe": "MFE_DOCKER_IMAGE",
}

def get_image_string(target_name):
    config = get_tutor_config()
    config_key = IMAGE_KEYS[target_name]

    print(config.get(config_key,"test"))
    return config.get(config_key)

def generate_backup_name(image, tag_suffix):
    """Safely creates a backup tag name, preventing invalid Docker syntax."""
    if ":" in image.split("/")[-1]:
        # If it already has a tag (repo:20.0.1), make it repo:20.0.1-BACKUP
        return f"{image}-{tag_suffix}"
    else:
        # If no tag is present, make it repo:BACKUP
        return f"{image}:{tag_suffix}"

# --- COMMAND: BACKUP ---
@click.command()
@click.argument("target", type=click.Choice(list(IMAGE_KEYS.keys()) + ["all"]))
@click.option("-t", "--tag", default="BACKUP", help="Tag suffix (default: BACKUP)")
def backup(target, tag):
    """Tag current Docker images as backups before pulling/building new ones."""
    targets = list(IMAGE_KEYS.keys()) if target == "all" else [target]
    
    for t in targets:
        image = get_image_string(t)
        backup = generate_backup_name(image, tag)
        click.echo(f"Creating backup: {image} -> {backup}")
        subprocess.run(["docker", "tag", image, backup], check=True)
        
    click.echo("Backup process complete.")

# --- COMMAND: RESTORE ---
@click.command()
@click.argument("target", type=click.Choice(list(IMAGE_KEYS.keys()) + ["all"]))
@click.option("-t", "--tag", default="BACKUP", help="Tag suffix to restore from")
def restore(target, tag):
    """Restore a broken image by reverting to the backup tag."""
    targets = list(IMAGE_KEYS.keys()) if target == "all" else [target]
    
    for t in targets:
        original = get_image_string(t)
        backup = generate_backup_name(original, tag)
        click.echo(f"Restoring backup: {backup} -> {original}")
        # Retag the backup as the primary image name
        subprocess.run(["docker", "tag", backup, original], check=True)
        
    click.echo("Restore complete. You may need to run `tutor local restart`.")

# --- COMMAND: CLEANUP/REMOVE ---
@click.command()
@click.argument("target", type=click.Choice(list(IMAGE_KEYS.keys()) + ["all"]))
@click.option("-t", "--tag", default="BACKUP", help="Tag suffix to remove")
def prune(target, tag):
    """Remove backup images to free up disk space."""
    targets = list(IMAGE_KEYS.keys()) if target == "all" else [target]
    
    for t in targets:
        image = get_image_string(t)
        backup = generate_backup_name(image, tag)
        click.echo(f"Removing backup image: {backup}")
        # check=False prevents the script from crashing if the backup doesn't exist
        subprocess.run(["docker", "rmi", backup], check=False)
        
    click.echo("Cleanup complete.")

# Register all commands with Tutor
images_command.add_command(backup)
images_command.add_command(restore)
images_command.add_command(prune)

Additional context

• I am not an experienced developer yet so I might be missing something about the whole docker images workflow
• I maintain custom cloned MFEs (private) and that's what I mean by "developing them"
• I am not sure if that is a niche problem or something that would be appreciated by other tutor site operators hence this issue
• Since thanks to the flexibility of tutor this solution may be implemented as a plugin, maybe it's not worth merging with base tutor
• I already use a plugin that links my custom registry by leveraging the IMAGES_PULL and IMAGES_PUSH hooks.
  • https://docs.tutor.edly.io/reference/api/hooks/catalog.html#tutor.hooks.Filters.IMAGES_PULL
  • https://docs.tutor.edly.io/reference/api/hooks/catalog.html#tutor.hooks.Filters.IMAGES_PUSH
• I know some Tutor plugins add new images to IMAGES_BUILD_REQUIRED (i.e. Aspects) so IMAGE_KEYS from my implementation would need to be linked with IMAGES_BUILD_REQUIRED somehow?
• Code is illustrative only as it can't be run directly. It uses my util function which I would say isself explanatory (just a wrapper around tutor's config so I don't have to find the root everytime)

If this is something worse pursuing I'd be happy to create a PR after possibly some implementation suggestions.

AI Declaration

Plugin created w the help of gemini 3.1 Pro