TODO

[jmartin-sul]

brain dump, will shape up into individual tickets as-is warranted, just want to move a possibly useful TODO list out of my personal notes:

• test on dlss-dev-suntzu (as dor user)
• add burndown box notes to README
• finish adding fedora settings to prod shared configs
  • host
  • proper cert location
• put current list of pids from sul-dor-prod on dlss-dev-suntzu (in the cocina-food-inspector deployment dir, see file starting with all_pids_, use whatever's latest)
  • load druid list into dlss-dev-suntzu "prod" database
• initial sketch of codebase
  • a service class that hits DSA object show endpoint to retrieve cocina representation of object
    • retrieves cocina representation for a druid
    • shunts output to success or failure dir depending on whether response is 200 OK or not
    • logs result of attempt
    • ability to pull object directly from fedora also (Dor.find using dor-services gem). flag to indicate whether that is done always, never, or only if cocina retrieval fails? default to last of those options? (include dor-services version when logging that this was done, also how to serialize?)
  • a model class that represents a DB record for a druid -- unique by druid, requires prefix, other fields: last_attempted_cocina_retrieval, last_attempted_fedora_retrieval, cocina_outcome (success/fail), fedora_outcome (success/fail)
  • a model class that represents a DB record for an attempt to retrieve a druid -- could then get rid of all the non-druid fields in the druid model and have attempted_on, cocina_outcome, fedora_outcome
  • an ActiveJob class that takes a druid and uses the service class to do a retrieval and write the result to DB using model class (delayed_job is already installed)
  • some canned queries and code fragments in README for queueing up jobs, reading a druid list to add new ones to the DB, etc
• rotate logs. but keep old logs around for a while (duration of this testing?), since they will be helpful in our analysis work. after loading fedora's current druid list into cocina-food-inspector's DB on sul-dev-suntzu, the prod log is 1.2 GB.
• would be nice to make this more of a real app. deploy to real VM, use shared configs for real, use PG instead of sqlite, sidekiq instead of delayed_job, extract DSC method per TODO, add rubocop, add tests, add badges (priority order?) (turn this into a real app #4)