Simplify ESQL specific operations using new ESQL runner #484
Conversation
|
In my local testing of these queries on a 5% ingest using |
The only new query remaining is esql_stats_enrich_control which runs as fast as all previous enrich queries, presumably because it groups by an existing indexed field. All the rest of the stats_enrich queries group by an enriched field, which presumably prevents some kind of optimization from being used, because they take literally 100x the time to execute. Locally the control take about 30ms, while the others take between 3s and 9s.
|
Additional local testing on 5% ingest shows that all existing So I've now removed all 7 slow queries from the schedule. I'll discuss with the ESQL team how best to handle this. One option is to bring back just one slow query, give it a very low iteration rate, and very long timeout. If this performance problem can be fixed, then we can bring back the remaining 6. |
* Simplify ESQL specific operations using new ESQL runner * Removed new stats_enrich queries that take long and cause timeouts The only new query remaining is esql_stats_enrich_control which runs as fast as all previous enrich queries, presumably because it groups by an existing indexed field. All the rest of the stats_enrich queries group by an enriched field, which presumably prevents some kind of optimization from being used, because they take literally 100x the time to execute. Locally the control take about 30ms, while the others take between 3s and 9s.
Also removed timeout since it was previously only used for _search queries which often took much longer. And made the pragma body a one-liner for easier reading.