Here’s a structured, interesting guide tailored for a database technician using Stellar Repair for MS SQL . It moves beyond basic “click-to-repair” and focuses on recovery strategy, forensic analysis, and workflow integration.
The Technician’s Deep Dive: Stellar Repair for MS SQL Why This Isn’t Just Another “Next→Next” Tool Most repair tools give you a recovered table. Stellar, when used right, gives you diagnostic intelligence — why the corruption happened, which objects are unrecoverable, and how to inject data with minimal downtime. Phase 1: Pre-Recovery Forensics (Don’t Open the MDF Yet) Technician step: Before loading the corrupt MDF, check:
Last known good backup (time + LSN) DBCC CHECKDB output (even if it fails, capture error numbers) sys.databases → state_desc , recovery_model
Why: Stellar’s scan mode depends on whether you have a page-level corruption (e.g., 823/824) vs. metadata corruption (e.g., 5173, 5171). Phase 2: Smart Scanning — Not All Modes Are Equal | Mode | Best for | Speed | Completeness | |------|----------|-------|---------------| | Quick | Corrupt non-clustered indexes, single page errors | Seconds | Low | | Standard | Missing system tables, partial data loss | Minutes | Medium | | Deep Scan (technician’s choice) | Severe corruption, dropped objects, truncated tables | Hours | High (raw page parsing) | Pro tip: Run Standard first, export recoverable data, then run Deep Scan only on specific object IDs if you suspect critical tables are missing. Phase 3: Selective Object Recovery (Not Just “All Tables”) Most techs ignore the tree view — don’t. stellar repair for ms sql technician
Expand → check Table Name , Rows , Data Consistency column Use Filter by:
Schema only Date range (if timestamps exist) Specific corrupted pages (if you know page IDs from SQL error log)
Recover as:
New MDF (clean schema + data) Direct to live DB (requires single-user mode on target) CSV/JSON for ETL into warehouse
Phase 4: Post-Recovery Validation Scripts (Your Signature as a Tech) After recovery, never trust “repair successful” message. Run these against the recovered DB: -- 1. Row count parity with known good backup (or approximate) SELECT COUNT(*) FROM suspect_table; -- 2. Foreign key sanity (orphaned rows after repair) SELECT fk.name, OBJECT_NAME(fk.parent_object_id) FROM sys.foreign_keys fk WHERE NOT EXISTS (SELECT 1 FROM sys.tables t WHERE t.object_id = fk.referenced_object_id); -- 3. Check for recovered NULLs in non-nullable columns SELECT COUNT(*) FROM suspect_table WHERE critical_column IS NULL AND critical_column IS NOT NULL;
Phase 5: Zero-Downtime Integration for Production Instead of restoring the entire corrupt DB: Here’s a structured, interesting guide tailored for a
Recover only corrupt objects into a staging MDF Use INSERT INTO ... SELECT with IGNORE_DUP_KEY (if indexes exist) Rename original → _corrupt_YYYYMMDD Rename recovered → original DB name Run DBCC CHECKDB with PHYSICAL_ONLY first, then full
When to NOT Use Stellar Repair