sp_IndexCleanup: performance + bug fix #721#723
Merged
erikdarlingdata merged 6 commits intomainfrom Mar 25, 2026
Merged
Conversation
…scans The LEFT JOIN to dm_db_index_usage_stats was executed via nested loops, scanning the full DMV per row (52M actual rows, ~4 minutes). Now pre-stages the DMV for the current database into #usage_stats with a clustered PK on (object_id, index_id) before the main query, then LEFT JOINs to the temp table instead. Handles the database cursor loop with TRUNCATE on re-entry. Tested on SQL2022 (StackOverflow2013) and SQL2016 (StackOverflow2010). Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…sage-stats sp_IndexCleanup: stage dm_db_index_usage_stats to fix 4-minute scan
Adds EXISTS filter against #filtered_objects when populating #usage_stats, so only indexes on objects we're analyzing are staged instead of every touched index in the database. Tested on SQL2022. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…sage-stats sp_IndexCleanup: scope usage stats staging to filtered objects
…ULL schema join Issue #721: Unique constraints were excluded from both sides of the Key Subset rule, so a NC index that was a prefix of a UC's keys was never flagged. Removed the NOT EXISTS check on the wider (target) side so UCs can serve as the superset target. The narrower-side check stays to prevent disabling UCs themselves. Schema join: Replaced LEFT JOIN sys.tables + LEFT JOIN sys.views + ISNULL(t.schema_id, v.schema_id) with a single JOIN sys.objects in all 4 query locations. Eliminates the non-SARGable ISNULL predicate that caused late filtering in the optimizer. Also scoped #usage_stats staging to #filtered_objects. Tested on SQL2022 (repro from #721 confirmed fixed) and SQL2016. Closes #721 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…et-and-schema-join sp_IndexCleanup: fix #721 unique constraint subset + replace ISNULL schema join
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Performance: Stage
dm_db_index_usage_statsinto temp table with clustered PK, scoped to#filtered_objects. Eliminates 52M-row nested loop DMV scan (~4 minutes → seconds).Schema join: Replace
LEFT JOIN sys.tables + sys.views + ISNULL(t.schema_id, v.schema_id)withJOIN sys.objectsin all 4 queries. Eliminates non-SARGable late filter.Bug fix (#721): Unique constraints now recognized as superset targets in Key Subset detection. NC indexes that are a prefix of a UC's keys are correctly flagged for DISABLE.
Closes #721
🤖 Generated with Claude Code