Index Analyzer
functionalReal index usage counters, least-read first, with a script-first DROP INDEX for the ones you choose.
Real usage counters for every index in the database, sorted least-read first, so the indexes nobody is using rise to the top. The columns are the engine's own numbers: seeks, scans, lookups, and updates from SQL Server's usage DMV; scans, tuples read and fetched, and on-disk size on PostgreSQL; read and write events from MySQL and MariaDB's Performance Schema; access and execution counts from Oracle's index usage tracking. A footer reminds you of the catch every DBA eventually relearns: counters reset on server restart, so a zero-read index on a freshly bounced server proves nothing yet.
Removing one is script-first: Script DROP INDEX prepares a commented,
reviewed DROP INDEX in a query tab for the row you selected. Indexes that
enforce a primary key or unique constraint are refused outright, with a note to drop the
constraint instead, the tool will not help you saw off a limb by accident.

Finding an index to drop
- Right-click the database and choose Tools › Index Analyzer... (or Index Analyzer... in the command palette for the current database).
- Look at the top rows. On SQL Server the order is fewest total reads first, and among those the most-updated first, an index that is never read but written on every insert is pure cost.
- Check how long the server has been up before trusting a zero. A month of real traffic means much more than an hour.
- Select the row and press Script DROP INDEX. The dialog closes and a new query tab holds the statement:
-- Review before running: this permanently drops the index and cannot be undone without rebuilding it. -- Usage counters reset on server restart — low numbers may only mean a recent restart. DROP INDEX [IX_Orders_LegacyStatus] ON [dbo].[Orders];
PostgreSQL gets DROP INDEX "schema"."index";, MySQL and MariaDB
DROP INDEX `index` ON `schema`.`table`;, and Oracle
DROP INDEX "OWNER"."INDEX";. Names are quoted for each engine, so odd
characters are safe.
Rebuilding instead
Script REBUILD works the same way for an index you want to keep:
ALTER INDEX ... REBUILD on SQL Server and Oracle, REINDEX INDEX on
PostgreSQL, headed by a comment that rebuilding takes locks and can run for a long time.
MySQL and MariaDB have no single-index rebuild, so the button explains that instead.
Columns by engine
| Engine | Counters shown | Sorted by |
|---|---|---|
| SQL Server / Azure SQL | type_desc, user_seeks, user_scans, user_lookups, total_reads, user_updates, last seek and scan times | Total reads, then updates (most first) |
| PostgreSQL | index_scans, idx_tup_read, idx_tup_fetch, size_mb | Scans, then size (largest first) |
| MySQL / MariaDB | read_events, write_events, total_events (system schemas excluded; up to 500 rows) | Reads, then writes (most first) |
| Oracle Database | index_type, total_access_count, total_exec_count, total_rows_returned, last_used, status (with an is_unusable flag), uniqueness, last analyzed | Access count, then table and index name |
Every engine also returns schema_name, table_name,
index_name, and is_enforcing, which the script buttons read. On
SQL Server, PostgreSQL, and Oracle, is_enforcing covers primary-key and unique
indexes; on MySQL and MariaDB only the PRIMARY index is flagged, so review a
unique index's drop script with that in mind.
Safety and permissions
- The pane only reads counters. Dropping and rebuilding happen only when you run the scripted statement yourself.
- On Oracle, the analyzer lists the current schema's indexes and needs Oracle 12.2 or
later plus
SELECTonDBA_INDEX_USAGE(orSELECT ANY DICTIONARY). Without it, the pane shows the error with that hint. Oracle samples its usage tracking, so the counts are approximate. - MySQL and MariaDB need the Performance Schema enabled for the counters to exist.
Available on SQL Server, PostgreSQL, MySQL/MariaDB, and Oracle Database (usage from DBA_INDEX_USAGE), whose engines keep the counters. SQLite, libSQL / Turso and DuckDB keep no index usage counters, so the tool says so there rather than showing zeros. ClickHouse keeps none either, and the menu points you at the index analysis in an estimated plan (the parts and granules each index pruned) instead. Redis has no indexes; every key is looked up by name.