"Migrations recently applied, need to reindex elastic" cron job

[jtimpe]

Description:

As part of #2870, we discussed the approach of running elasticsearch reindexing on a cron schedule to avoid any end-user impact. This approach is imperfect: reindexing only needs to be done when migrations are added to a particular set of models (search_indexes.models), and even then only for that model. Reindexing on any cron schedule is inefficient and leaves the door open to a failure or partial success when we don't need to take that risk.

To address this, create another cron or post-deploy task that

• Checks if a reindexing is needed by checking if migrations to any search_indexes.models have been recently applied
• and, either:
  • Publishes a flag to indicate whether a reindex is required. Update #2881 to use this flag to determine whether or not to perform a reindex. A "better" implementation would include the ability to determine individual models migrations have been created for and only reindex those tables.
  • or; notifies a system administrator that a manual reindex is required

Acceptance Criteria: Create a list of functional outcomes that must be achieved to complete this issue

• [ ] Create the cron or post-deploy task to update a "needs reindexing" flag
• [ ] Create a notification OR update #2881 to utilize to flag created by the previous task
• [ ] Testing Checklist has been run and all tests pass
• [ ] README is updated, if necessary

Tasks: Create a list of granular, specific work items that must be completed to deliver the desired outcomes of this issue

• [ ] Discuss cron or post-deploy task for flag-setting
• [ ] Discuss manual vs automated reindexing (notification vs cron)
• [ ] Create the celery task for flag-setting
• [ ] Create the notification or update existing cron PR for reindexing
• [ ] Run Testing Checklist and confirm all tests pass

Notes: Add additional useful information, such as related issues and functionality that isn't covered by this specific issue, and other considerations that will be helpful for anyone reading this

• PR #2881 implements a cron job for elastic reindexing
  • including usage of --parallel, --use-alias flags
  • routes all log output to /dev/null - nothing useful is generated by the library logs, silencing all log output is a large performance gain
  • includes a 10s wait between each model being reindexed - to allow elastic instance to "recover" in between sets

Open Questions: Please include any questions or decisions that must be made before beginning work or to confidently call this issue complete

• Cron or post-deploy task?
• Manual vs automated reindexing (notification or cron)?

[robgendron]

Backlog Refinement:

Manual vs automated reindexing (notification or cron)? Automated

[robgendron]

May no longer be the solution - need to verify.