Overview
In this article, we will define the Activity Engine processing queues that are relevant to the Recommender service and their expected values. Knowing this is essential in troubleshooting as it gives you information about the items that are awaiting processing by the activity engine node and helps you identify any unusual process behaviors.
Information
The queue depths displayed on the Activity Engine page of the Admin Console show the number of items awaiting processing by the Activity Engine node. When an activity is sent to the Activity Engine, the activity is queued and processed as quickly as possible.
Each queue, or group of queues, serves a specific purpose:
If you see an R after a queues' name (e.g., such as in the Command or Shredder queues) it means that the queue in question is used (besides EAE processing) also for Recommender service processing.
The following are the queues (also) used by the Recommender service:
- Command
Used when specific tasks need to be scheduled and run on the Activity Engine node, such as a user rebuild or Recommender refreshed, ideally, the depth shows remain as close to 0 (zero) as possible.
- Recommendation
After requesting recommendations, the queue is used to check for and process the results (does not affect normal EAE processing). Should the queue grow and never drop, it would indicate a communication problem with the Recommender service.
- RecommendationRequest
We issue recommendations requests from the Recommender service and then cache them locally. This queue contains pending requests. Ideally, this queue depth should remain low.
- Shredder
If Jive Genius is enabled, this queue will fill up with items as activity comes in and needs to be shredded before being sent to Jive Genius. Depth should remain 0 unless a reshred is requested.
- Reshred
This queue will reprocess the data to send to the Recommender service. Typically, it would be used if the recommender gets out of sync and the depth should remain at 0 unless a reshred is requested.
Comments
0 comments
Article is closed for comments.