Local databases and files

Tested with: zb migrate 0.1 · pg_dump 17 · mysqldump 8.4 · valkey-cli 8.1 · mongodump 100.10 · sqlite3 3.46 · duckdb 1.3

--from local runs the engine’s own dump tool on your machine and streams it to a pre-signed upload URL; a file path uploads the file as-is. Nothing needs to be reachable from the internet.

# Local server (uses PG* env vars or --host/--port/--user)
zb migrate --from local --engine postgres --db myapp --create

# Existing dump (custom or plain format, optionally gzipped)
pg_dump -Fc myapp > myapp.dump
zb migrate --from ./myapp.dump --engine postgres --create
zb migrate --from ./myapp.sql.gz --engine postgres --to inst_abc

zb migrate --from local --engine mysql --db myapp --create

mysqldump --single-transaction --routines --triggers --set-gtid-purged=OFF myapp | gzip > myapp.sql.gz
zb migrate --from ./myapp.sql.gz --engine mysql --create

zb migrate --from local --engine valkey --create          # runs valkey-cli --rdb against localhost:6379

valkey-cli --rdb dump.rdb                                  # or an existing RDB snapshot
zb migrate --from ./dump.rdb --engine valkey --create

zb migrate --from local --engine ferretdb --db myapp --create    # runs mongodump

mongodump --db myapp --archive=myapp.bson.tar
zb migrate --from ./myapp.bson.tar --engine ferretdb --create

zb migrate --from ./app.sqlite --engine libsql --create
zb migrate --from ./app.db --engine libsql --to inst_abc --drop-target

WAL-mode databases are checkpointed on upload; keep the -wal file next to the database until the upload is done.

zb migrate --from ./analytics.duckdb --engine duckdb --create --size s1
zb migrate --from ./events.csv --engine duckdb --include events --to inst_abc
zb migrate --from ./orders.parquet --engine clickhouse --include orders --to inst_abc
zb migrate --from ./metrics.csv --engine questdb --include metrics --to inst_abc

CSV and Parquet files load into one table, named with --include. DuckDB takes .duckdb files and CSV; for Parquet, load it into a .duckdb file first (duckdb app.duckdb -c "CREATE TABLE t AS FROM 'file.parquet'"). QuestDB takes CSV only. See DuckDB, ClickHouse and QuestDB.

If the target engine is not orderable yet (its engine page shows the status), the file formats are accepted now and queued until it is.

Size limits and resumable uploads

Uploads are S3 multipart with resume: an interrupted zb migrate continues where it stopped. Files up to the target size’s storage ceiling are accepted; --create picks a size whose ceiling fits the file with 30% headroom.