During the initial snapshot import of the SNOMED CT RF2 file there are up to 10K Elasticsearch scroll contexts open at once. This may be slowing down the import. We should investigate if these search contexts can be closed more quickly in code, perhaps in the spring-data-elasticsearch layer, rather than relying on the scroll context timeout which is likely at the moment.
During the initial snapshot import of the SNOMED CT RF2 file there are up to 10K Elasticsearch scroll contexts open at once. This may be slowing down the import. We should investigate if these search contexts can be closed more quickly in code, perhaps in the spring-data-elasticsearch layer, rather than relying on the scroll context timeout which is likely at the moment.
See the
open_contexts
stat when importing here http://localhost:9200/_nodes/statsThanks to @rorydavidson for finding this.