operator wisdom
7 traps
Dialect Traps
Completeness isn't only row counts — it's knowing where the warehouse and the Redshift Data API will quietly hand you a wrong answer. These are the things we learned so you don't get quietly wrong answers. Click any trap to read the detail.
3
correctness
2
modeling
2
hygiene
svv_all_tables matches what you can actually SELECT from — prefer it. The disagreement itself is worth asserting on.
Snapshot tables — never query them as current
Migration artifacts that sit right next to live tables. There are 13 of them across the warehouse; wsl carries 10 — half its schema. Match these globs and exclude them when you enumerate a schema's tables, or your counts silently double.
- *_retired_YYYYMMDD
- *_preclean_YYYYMMDD
- *_ghosts_*
- *_seed_*
Pulled with AWS_PROFILE=asd against Redshift Serverless (us-west-1 / default-workgroup / dev).