v0.99.16

Multi-database MySQL migration and continuous sync. A single sluice run can now connect to a MySQL server and migrate — and continuously sync — many databases at once, each landing in its own same-named target namespace, analogous to how a Postgres source carries multiple schemas. Drop-in upgrade from v0.99.15 — purely additive; without the new flags, every existing single-database run is byte-identical.

Added

• migrate across multiple MySQL databases in one run. New flags --include-database <glob> / --exclude-database <glob> (mutually exclusive, repeatable) and --all-databases. When any is set, the source DSN is a server connection (its database component is optional), sluice enumerates the server's databases, and each selected database is migrated to a same-named target namespace — a Postgres schema (MySQL→Postgres) or an auto-created target database (MySQL→MySQL, via CREATE DATABASE IF NOT EXISTS).

  System databases (information_schema, performance_schema, mysql, sys) are always excluded. Cross-database foreign keys are preserved when both databases are in scope (applied in a final pass once every database's tables exist); a foreign key pointing at a database outside the selected set is refused loudly — sluice can't guarantee the referent exists on the target, so it never silently flattens the reference.

• sync start across multiple MySQL databases — cold-start, CDC, and resume. The same --include-database / --exclude-database / --all-databases flags on sync start give continuous multi-database replication:

  • One consistent snapshot across all databases. The cold start captures a single START TRANSACTION WITH CONSISTENT SNAPSHOT on one pinned connection and one binlog position, so the snapshot→CDC handoff is a single gapless cut across every selected database.
  • One CDC stream, routed per change. Steady-state CDC rides the server-wide MySQL binlog as one stream and routes each change to its source database's target namespace — not N streams to coordinate.
  • Warm resume. A stopped stream resumes from the one persisted server-wide position without re-copying.

  Works MySQL→MySQL and MySQL→Postgres.

Fixed

• Silent-loss boundary gap in the binlog snapshot→CDC handoff (single- and multi-database). The binlog snapshot opener captured the row view (START TRANSACTION WITH CONSISTENT SNAPSHOT) and the CDC start position (SHOW BINARY LOG STATUS) as two separate statements. A transaction committing in the window between them landed in neither the snapshot nor the CDC tail — a silently lost row. The capture is now wrapped in FLUSH TABLES WITH READ LOCK … UNLOCK TABLES (the mydumper/Debezium consistent-snapshot pattern), so the snapshot view and the binlog position name the exact same logical cut. FLUSH TABLES WITH READ LOCK needs the RELOAD privilege; absent it, sluice warns and falls back to the prior lock-free capture rather than failing. The fix lands in the shared snapshot opener, so it closes the gap for both the new multi-database cold start and the pre-existing single-database binlog snapshot path.

Compatibility

• No breaking API or CLI changes. Drop-in from v0.99.15. Multi-database mode engages only when a --*-database / --all-databases flag is set; without them, single-database migrate and sync start are byte-identical (same snapshot, same position, same apply path). This is MySQL-source fan-out — PlanetScale/VStream multi-keyspace and the reverse Postgres-source→MySQL-multi-database direction are tracked follow-ons. A MySQL→MySQL multi-database target DSN must name a "home" database for the sync control table (it errors clearly if absent); per-source user data still routes to its own database.

Who needs this

• Anyone consolidating or replicating a MySQL server with many databases (the classic "connect as root, copy everything" case). One sluice migrate --all-databases or sync start --all-databases moves/keeps-in-sync the whole server, each database in its own target namespace — instead of one process per database.
• MySQL→Postgres movers get each MySQL database as its own PG schema automatically, no per-database --target-schema juggling.

---

Install: brew install sluicesync/tap/sluice · go install sluicesync.dev/sluice/cmd/sluice@v0.99.16 · Container: ghcr.io/sluicesync/sluice:0.99.16
Full changelog: https://github.com/sluicesync/sluice/blob/main/CHANGELOG.md