Failure when rematching a large number of lists

[adam-collins]

When using the admin button rematch all, it can fail on smaller machines when there are a large number of lists. This should not happen. There is a workaround implemented but this workaround is not ideal as it involves restarting tomcat.

During a rematch all, existing name matches are first removed, then items are iterated, with batches sent to namematching service.

For this issue, make changes to rematch all.

• Perform an update to existing matches only after a diff. This is instead of the removal of matches at the beginning of the rematch all.
• Use smaller transaction boundaries and wait for the database flush to complete.
• Enable parallel updates with a configurable number of threads.
• Iterate over lists before individual list items. This is instead of iterating through list items only.

[qifeng-bai]

288

[qifeng-bai]

Q: Iterate over lists before individual list items. This is instead of iterating through list items only. A: nameExplorerService iterates over species items, not lists

Q:Perform an update to existing matches only after a diff. This is instead of the removal of matches at the beginning of the rematch all. A: Got answer from Simon Sherrin and Mahmoud: the taxonConceptID is changed when we updated the Taxonomic Backbone, yes.The general rule is - you can't rely on taxonConceptIDs to be persistent across updates to the Taxonomic Backbone. (BIE reindex)

In this case, the taxonConceptID will be changed anyway. When we are rematching, we don't have to do a diff, just need to update directly