Baseline capture for CockroachDB to YugabyteDB parity checking

Shipped

9 ஆகஸ்ட், 2026 அன்று 9:41 PM UTC

Author

Kamo

Commit

e93b355

Two passes: cheap batched count(*), then an order-independent checksum. The checksum sums per-row md5 hashes as integers rather than the obvious md5(string_agg(h ORDER BY h)). Both are order-independent, but string_agg materialises ~186 MB for geolite_blocks' 5.8M rows; it ran over 2.5 minutes on a 330K-row table before being cancelled. The sum() form is constant memory and does the whole database in 1m17s.