sgup432 opened a new issue, #15097: URL: https://github.com/apache/lucene/issues/15097
### Description We recently discovered a bug in OpenSearch where LRUQueryCache was taking ~80% of the heap, though it is capped at 10%(of heap) from OpenSearch side. While analyzing heap dump, we found out that there were bunch of large Boolean queries still lingering around in the cache bloating its size beyond its max size. While looking at the LRUQueryCache [code](https://github.com/apache/lucene/blob/main/lucene/core/src/java/org/apache/lucene/search/LRUQueryCache.java#L428-L433), it seems like we calculate the query size via `accountableQuery.ramBytesUsed()` only if the desired query implements `Accountable` interface and provide its own implementation. In this case, BooleanQuery doesn't provide such implementation and we default to 1kb, leading to underestimation of query size causing the cache size to bloat up with such large queries. Below is a snapshot from MAT where it shows LRUCache holding ~23gb of heap in an instance where we had 32gb of heap, and LRUCache is configured to be ~3.2gb in size in this case. And majority of being held by ~6k BooleanQuery each around ~3.2mb in size. <img width="1042" height="137" alt="Image" src="https://github.com/user-attachments/assets/75d6403a-dfd6-4562-8510-ff454d899895"; /> ### Solution Below are the solutions I can think of: 1. I think we can make BooleanQuery to implement Accountable interface and calculate rough estimate of ramBytesUsed from its desired clauses? Still better than using 1kb as default. 2. Or Disable caching BooleanQueries with very high number of clauses. Though this still won't be accurate enough. 3. Or have a different way to calculate cache's key, value size. I don't way an exact way to do this at this point. ### Version and environment details _No response_ -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
