Term Suggestions Synchronization
This page explains how and when to synchronize and recompute Term Suggestions. For more information on Term Suggestions functionality, see Term Suggestions: Behind the Scenes.
In versions 13.8 and earlier, Term Suggestions and the Metadata Management Module (MMM) communicated using a term suggestions database, which needed to be configured. Since version 13.9, the connection is enabled via Api service and it is no longer necessary to configure the database connection. As Term Suggestions services have improved in stability, you no longer need to initiate Term Suggestions synchronization from the application. The only exception to this is when modifying the metadata model, e.g. by restoring the database from a snapshot or by importing any initial content (default, demo, etc.).
Synchronization
To perform the synchronization, do the following:
-
Go to Global Settings > Term Suggestions tab:
-
Click the Synchronization button to initiate the synchronization process.
The synchronization can have one of the following statuses:
-
Synchronization has finished successfully, no resynchronization is in progress.
-
Resynchronization is in progress.
-
Terms Suggestions AI services are not available due to a connection failure or because they are not running. For more information, see the Deal with errors in synchronization section.
-
-
Wait for the synchronization to finish. When the synchronization becomes available again, it indicates that the synchronization was completed successfully. When the synchronization process is completed, the information in Term Suggestions AI services becomes up-to-date with the metadata, and the term suggestions recomputation process can begin.
It might take up to a few hours for all the suggestions to be recomputed and refreshed, depending on the number of profiled attributes in the whole catalog. The following table contains the approximate time for recomputing term suggestions for your number of attributes:
Number of Attributes Time for term suggestions recomputing 1 000
< 2 sec
10 000
< 12 sec
50 000
~ 1 min
100 000
< 3 mins
150 000
~ 5 mins
200 000
< 8 mins
250 000
~ 10 mins
300 000
< 14 mins
350 000
< 17 mins
1 000 000
~ 1.5 hours
2 000 000
~ 6 hours
3 000 000
~ 11 hours
When working with a large catalog, that contains ~15,000,000 published attributes, it is necessary to do the following to ensure the correct work of term suggestions synchronization:
-
Make sure that the
ataccama.one.term-suggestions.neighbors.cache.attributes-limitis increased to the appropriate one. For more information, see neighbors-configuration.adoc. -
Increase the
ataccama.one.apyc.heartbeat_timeoutto a big enough number (for example, 3600s) to allow the service to load the cache. For more information, see recommender-configuration.adoc.
-
Deal with errors in synchronization
It is possible that the synchronization is not available for some reason other than being in progress.
This can happen when Term Suggestions services cannot connect to the MMM, which means the services are completely unavailable. In that case, the connection will be fixed automatically in a few minutes. However, it might be necessary to debug and resolve the connection issue with the help of an admin who has access to the infrastructure, monitoring, logs, etc.
Was this page useful?