This is an exploration story pursuant to last planning meeting. The hypothesis is: the closer we can make the mediawiki/extensions/skin provisioning tooling in the chart resemble patchdemo's provisioning tooling, the easier it will be to implement/port the long tail of provisioning/configuring features from patchdemo to catalyst.

There are three two options for vendoring this script into initialisation containers:

• create and publish a new image with the scripts copied in via blubber or docker build
  • (+)this is similar to what mediawiki does for it's local development docker environments
  • (-) we do not recommend this approach, as it will make the feedback loop longer (i.e. you'll have to recreate the image any time the tooling changes
• have helm inject these scripts and then call them in init containers
  • (+) this allows the scripts to be includes as individual files (and thus can be tested individually)
  • (+) changing scripts only requires updates to the chart and not the underlying image
  • (-) injecting scripts through helm is a little awkward
  • (-) the scripts will not have access to {{.Values}}, so we'll need to figure out the best way to invoke these scripts with the proper configuration
• make the scripts helm templates and include the rendered template in the init container
  • • (+) these templates _do_ have access to {{.Values}}
    • (-) these template files are not _pure_ bash/shell/etc, and may render weird in IDEs
    • (-) the interface between the scripts and {{.Values}} is not defined, i.e. if the {{.Values}} schema changes, it may not be immediately obvious was needs to change in the script templates
  • overly complicated examples from prototype:
    • https://gitlab.wikimedia.org/repos/qte/catalyst/ci-charts/-/blob/extension-templates/mediawiki/templates/_wikilambda.tpl?ref_type=heads
    • https://gitlab.wikimedia.org/repos/qte/catalyst/ci-charts/-/blob/extension-templates/mediawiki/templates/deployment.yaml?ref_type=heads

Technical hints/notes:

• these scripts could be placed under a scripts folder in the helm chart
• these scripts can then be injected with a ConfigMap and a mounted Volume in the init-container
  • see this stackoverflow answer
  • this conversation with ChatGPT
  • we might be able to inject relevant configuration back setting them in environment variable or as arguments when invoking the script
  • keep an eye on the goal of reusing the scripts in patchdemo