Migration Plan

Plan for migrating from Oracle to PostgreSQL:

1. Scope the problem. Get a count of legacy workflow and legacy workflow_archive tables
   • On workflow-server-rails
     • ActiveRecord::Base.connection.query('SELECT count(*) from workflow_archive')
     • ActiveRecord::Base.connection.query('SELECT count(*) from workflow')
   • On workflow-archiver-job
     • DB = Dor::WorkflowArchiver.new.conn
     • DB.fetch('SELECT count(*) from workflow').first[:'count(*)'] (currently 572785)
     • DB.fetch('SELECT count(*) from workflow_archive').first[:'count(*)'] (currently 126,069,766)
2. Disable the workflow-archiver-job (remove from crontab) so that no new rows get added to legacy workflow_archive.
3. Export legacy workflow_archive to a CSV?
4. Import the CSV as workflows
5. Select all the unique druids in legacy workflow
6. For each druid, query dor-services-app to get the current version:
   • version = dor_conn.get("/v1/objects/#{druid}/versions/current").body
   • See https://github.com/sul-dlss/workflow-archiver/blob/5f0cb9b2282fb837b5b5accd72b72b3dc2882040/lib/dor/workflow_archiver.rb#L28
7. Then query all the rows matching that druid from the legacy workflow table, and insert them into the workflows table with the version we got from dor-services-app.
   • See https://github.com/sul-dlss/workflow-archiver/blob/5f0cb9b2282fb837b5b5accd72b72b3dc2882040/lib/dor/workflow_archiver.rb#L127-L139
8. Update the archive action from dor-services-app
   • I think this can be a no-op, and eventually be removed.
   • called from https://github.com/sul-dlss/dor-services/blob/517c9f147c51ec852fcc28c831ceddb860619a0d/lib/dor/services/merge_service.rb#L73)
   • https://github.com/sul-dlss/dor-services-app/blob/094f1dce407f6f6d5d4acecb95ff7d074c027d31/app/controllers/workflows_controller.rb#L8-L13
9. Update the sul-lyberservices-* proxy to point at the new workflow service.
10. Decommission workflow-archiver-job and deprecate workflow-archiver
11. Decommission old Java workflow service.