Changelog¶

Unreleased¶

No changes yet.

--as-of-system-time=follower: pin follower_read_timestamp() so exports are served by the nearest replica (follower reads), keeping the consistent-snapshot guarantee. Reads use default priority to stay low-impact. Fails fast with a clear message if the cluster lacks the follower-reads entitlement.

--as-of-system-time on export: read all table data at one pinned cluster timestamp for a consistent point-in-time snapshot. The bare flag pins cluster_logical_timestamp(); an explicit value (interval, timestamp, or decimal) is used verbatim. The pinned timestamp is recorded in each manifest as as_of_system_time.

Object naming is now three-part database.schema.table everywhere (filenames, manifests, resume-log keys, and --tables input). Pre-0.4.0 two-part dumps are not compatible with the 0.4.0 loader.
--tables two-part input now means schema.table (database taken from --db), not the legacy database.table.
Minimum supported Python is now 3.10.

Export and restore now work for objects in non-public schemas (the data exporter previously crashed on three-part names, and schema export silently dropped the schema component).
Mixed-case and reserved-word identifiers are now correctly quoted.
load honors --host/--port/--certs-dir (the psycopg2 path previously ignored them and only read CRDB_URL/localhost).
Schema export no longer duplicates DDL when the aggregate file already exists.
Schema load splitting handles semicolons inside string literals and function bodies (via sqlparse).
Selective data export (--tables) resolves the database from --db instead of misreading the first name segment as the database.
Replaced deprecated datetime.utcnow().

Native SHOW CREATE ALL TYPES / SHOW CREATE ALL TABLES for full-database dumps (dependency-ordered, with foreign-key constraints validated post-load).
Centralized identifier model (crdb_dump/utils/identifiers.py) used across all modules for naming and safe SQL quoting.
Regression tests for CockroachDB VECTOR data round-trips (SQL and CSV).

Dependencies upgraded to current GA releases (SQLAlchemy 2.0.51, sqlalchemy-cockroachdb 2.0.4, click 8.4.2, PyYAML 6.0.3, psycopg2-binary 2.9.12, boto3 1.43.36, sqlparse 0.5.5).
Data chunk files renamed from <table>_chunk_NNN.* to <db.schema.table>_NNN.*; manifests to <db.schema.table>.manifest.json.