fix(drizzle-kit): limit Postgres introspection fanout

[joe-thom]

Why

Datadog from repl-it-web's publish database-preparation flow shows prepareDatabaseMigrations failures during phase:introspect with pg.Pool.waitingCount in the hundreds/thousands. The immediate quick fix in #24 gates calls before pool.query(...), but the deeper root cause is in drizzle-kit itself: pgSerializer.fromDatabase() eagerly creates one async task per table/view.

For large user schemas that can create thousands of concurrent table/view introspection tasks. Each task performs multiple catalog queries, so the pg pool queue can grow until queued acquires hit connectionTimeoutMillis.

What changed

Adds optional Postgres introspection fanout limiting:

introspectPgDB(db, filters, schemaFilters, { tableConcurrency: 4 })

and threads the option through to pgSerializer.fromDatabase().

fromDatabase() now uses a small mapWithConcurrency() helper for the table and view/materialized-view introspection phases. Existing behavior is unchanged when tableConcurrency is omitted.

Test plan

• pnpm --filter @drizzle-team/drizzle-kit exec vitest run tests/api.test.ts tests/pgSerializer.test.ts passed
• pnpm --filter @drizzle-team/drizzle-kit run tsc passed
• pnpm --filter @drizzle-team/drizzle-kit run build passed
• pnpm lint passed

Revertibility

Fully revertible. This is an additive API option and preserves existing behavior when tableConcurrency is omitted.

Follow-up

After this stack is published, repl-it-web should bump @drizzle-team/drizzle-kit and use both guards in the publish DB-preparation flow:

const db = await preparePgDB(pool, { queryConcurrency: 4 });
const schema = await introspectPgDB(db, [], ['public'], { tableConcurrency: 4 });

[joe-thom]

• fix(drizzle-kit): limit Postgres introspection fanout #25 👈 (View in Graphite)
• fix(drizzle-kit): limit prepared Postgres query concurrency #24
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.