Web / Apache Lucene Interview questions
How do you troubleshoot slow queries in Lucene?
Slow-query investigation in Lucene usually follows a fairly consistent checklist rather than guesswork:
- Inspect the query shape - leading wildcards (
*term) and broad fuzzy queries are inherently expensive since they can't use the term dictionary efficiently. - Check field analysis - a mismatched or overly complex Analyzer can bloat the term dictionary and slow term lookups.
- Check segment count - too many small segments increases per-segment overhead; this usually points at MergePolicy or commit frequency issues.
- Warm the searcher - a freshly reopened IndexSearcher with cold caches (especially DocValues) will be slower on its first queries.
- Use IndexSearcher.explain() on a slow query to see exactly which clauses are contributing the most scoring work.
Most real-world slowdowns trace back to one of the first two items - an unbounded wildcard/fuzzy query, or an Analyzer producing far more distinct terms than the field actually needs.
Which query pattern is inherently expensive because it can't use the term dictionary efficiently?
Which method helps show exactly which clauses drove a query's scoring work?
More Related questions...
