-
Type: Improvement
-
Status: Resolved
-
Priority: Minor
-
Resolution: Fixed
-
Affects Version/s: None
-
Fix Version/s: 10.3
-
Component/s: Core, Events / Works, Streams
-
Impact type:Configuration Change
-
Upgrade notes:
-
Sprint:nxFG 10.3.5, nxFG 10.3.6
-
Story Points:3
FulltextUpdaterWork was created by NXP-10864 in order to write newly-extracted fulltext into the database from a single thread, to avoid overloading some databases. This is separate from the FulltextExtractorWork which does the conversion from the binary into text, but currently does not store this text.
Because the extracted text has to be passed to a second work, the text is stored in a queue for asynchronous processing. However this may be very large, easily several megabytes, which is a problem because queues like Kafka or Chronicle are not designed to work efficiently with such large volumes.
The original problem leading to this design was a database we rarely use (SQL Server) and in addition we have other means of solving the concurrency problem these days (retry on ConcurrentUpdateException).
—> merge back FulltextUpdaterWork into FulltextExtractorWork
- depends on
-
NXP-5689 VCS: store PostgreSQL fulltext in clear text
- Resolved
-
NXP-10864 Write to fulltext table single-threaded
- Resolved
-
NXP-8797 Make fulltext normalization algorithm configurable
- Resolved
- is related to
-
NXP-14672 Factor common code for VCS and DBS
- Resolved
-
NXP-17862 Improve fulltext extraction
- Resolved
-
NXP-25279 Make the raw binary text available for processing
- Resolved
-
NXP-26618 FulltextUpdaterWork class not found after migration
- Resolved
- is required by
-
NXP-25781 Allow per-field fulltext indexing for MongoDB
- Resolved
-
NXP-25359 FulltextExtractorWork should only get called once per blob
- Resolved
- Is referenced in