The first query to a completion suggester field after an indexation is taking several seconds to some minutes, whereas the following queries are in the tens of milliseconds (tested in Elastic 8.3).
I think this might be due to lazy loading of new FST segments into memory - is this possible?
Having a program periodically issue a query (e.g. every 10 seconds) resulted in much higher stability in the request latency, even while indexing is happening.
This seems like a valid use case for the index warmers, but unfortunately, they have been removed as of version 8.3.
Is there any alternative to having that periodic pinging? Should this be reported as an issue?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.