Read clustered recs from Vitess

Goal

We can't exclude enough seen posts with the current ElasticSearch solution. #2363 (closed) will enable us do a JOIN query in order to exclude seen results much easier, and with much less bandwidth usage.

What needs to be done

Read clustered results from MySQL.

CREATE TABLE clustered_entities_recs (
  cluster_id int,
  entity_guid bigit,
  score float,
  ...
  PRIMARY KEY (cluster_id, entity_guid)
)

ES is currently:

{
  'entity_guid': record[0],
  'entity_owner_guid': rprops[0],
  'cluster_id': record[1],
  'score': record[2],
  'total_views': rstats[2],
  'total_engagement': rstats[3],
  '@first_engaged': rstats[0].strftime("%Y-%m-%dT%H:%M:%SZ"),
  '@last_engaged': rstats[1].strftime("%Y-%m-%dT%H:%M:%SZ"),
  '@last_updated': update_str,
  '@time_created': rprops[1].strftime("%Y-%m-%dT%H:%M:%SZ")
}

QA

Ensure results are being returned.

UX/Design

N/A

Personas

N/A

Experiments

No, but do make a feature flag

Acceptance Criteria

• Schema for MySQL
• Querying results from Vitess and removing ElasticSearch
• Spec tests

Definition of Ready Checklist

• Definition Of Done (DoD)
• Acceptance criteria
• Weighted
• QA
• UX/Design
• Personas
• Experiments

Read