Issues

With big database, Sphinx offline indexing requires:

• Resource consumption: at least 4 CPU and 16 GB RAM
• Time: The bigger the data, the more it takes time to reindex
• Cron: a job needs to be scheduled to reindex

On OCP, the sphinx indexer is launched inside a Thread,
https://github.com/3scale/porta/blob/master/lib/tasks/openshift.rake#L38-L86

Meaning that whenever we deploy a pod, a full reindex is done

Dev notes

https://freelancing-gods.com/thinking-sphinx/v3/real_time.html

• Realtime indexing will remove the dependency on Cron job or the background thread
• Realtime indexing will allow asynchronous indexation by sidekiq to leverage the web unicorn worker.

Testing procedure

• system-sidekiq should be running
• system-app should be running
• system-sphinx should be running

Create or modify an Account or CMS page

Before: They can be searched after 30 minutes

After: They can be searched immediately after the sidekiq job SphinxIndexationWorker has been running.

h3. Upgrade procedure

From 2.8 to 2.9

Delete the content of system-sphinx-database volume
Redeploy the system-sphinx pod
Reindex launching the tak rake sphinx:enqueue