README

This is UO's exhibit app, built on top of Spotlight

Infrastructure

This application is expected to be run inside docker or podman containers for both development and production. compose.yml defines all the services necessary for the application to run.

Key services:

• proxy is a Caddy proxy that caches some static content for faster delivery.
• web is the actual Rails app being served up by Puma.
• sidekiq runs all background jobs.

Local dev

Configuration

It is highly recommended that you keep RAILS_ENV set to production. The development setup will make the edit loop faster, but so many things are different that it can cause problematic "drift" between prod and dev.

For instance, if you're trying to figure out a performance problem, all the caching that is only done in a prod environment will completely ruin any benchmarking or metrics. Not that I would forget this, of course, but I've heard from a friend that it can be annoying.

Getting data

To mirror production's database, you'll need to set up build/mirror-vars.sh (see the example file for variables that are needed) and then run build/mirror.sh. See mirror-vars-example.sh for details.

Note: this relies on having the podman rsync stuff installed on the podman server!

Making Solr aware of your data

Note: you must have your uploads available to the sidekiq service if you want your reindex to work! The index procedure will "lose" images and even some metadata if you run a reindex without properly mounting in uploads!

If you only rely on the volumes, you're fine: web and sidekiq both use the same volume for the uploads dir. But if you mount things in from an external source, you must mount the same things into both containers!!

---

To index an exhibit, you'll need to go into a rails console, pick the exhibit, and tell it to reindex "later":

# Very slow: reindex everything
Spotlight::Exhibit.find_each do |e|
  e.reindex_later
end

# Less slow: pick a single exhibit to reindex
Spotlight::Exhibit.find(<id>).reindex_later

Note that "later" is somewhat misleading. The jobs will be sent to sidekiq immediately, and executed as soon as sidekiq is able.

Then, making sure you have the sidekiq service running, go grab some coffee or something. With the prod data, this will take a few minutes.

Starting the stack

• Copy compose.override.example.yml to compose.override.yml and edit as needed. Do not use the copied override exactly as-is!
• Use docker compose, e.g., docker compose up -d proxy

Rails console stuff to know

The "base" object type appears to be a Spotlight::Resource. Exhibits are Spotlight::Exhibit. You can find an exhibit and its resources something like this:

e = Spotlight::Exhibit.find(5)
resource = Spotlight::Resource.find(e.resource_ids.first)

You cannot currently use e.resources because some resources in the database have a model that is no longer in the codebase. Decoupling code and data is way out of scope, but something to consider when we do major work on this again.