Database Maintenance

Use this skill when planning or performing operational maintenance on relational databases, pgvector stores, or modern lakehouses to preserve health, query performance, and financial efficiency.

When to Use

• compacting small files in Apache Iceberg or Delta Lake tables via bin-pack rewriting and merging Deletion Vectors to target 256MB Parquet blocks
• consolidating fragmented Avro manifests via rewrite_manifests for rapid query planning
• expiring stale lakehouse snapshots with 7-day TTL while preserving a 50-snapshot safety floor
• vacuuming orphaned data files with a mandatory 72-hour grace window to safeguard concurrent writers
• tuning multi-dimensional Z-Order clustering on high-cardinality columns and partition pruning strategies
• calculating Data FinOps metrics (file reduction ratio, storage reclaimed GB, monthly S3 GET savings) and tagging runs with CostCenter, Environment, and TableOwner
• monitoring analytical query costs, setting 50 GB scan limits, and enforcing warehouse auto-suspend (≤ 60s)
• rebuilding fragmented B-tree, HNSW, or IVFFlat pgvector indexes concurrently

Core Rules