apache/superset · error · CommandInvalidError

; ".join(str(message) for message in ex.messages)

Error message

; ".join(str(message) for message in ex.messages)

What it means

When a v1 dataset import assigns a dataset a non-default catalog while the target database has multi-catalog support disabled, the underlying MultiCatalogDisabledValidationError is re-raised as CommandInvalidError so the REST layer returns 422 (validation error) instead of a generic 500. The message is the joined validation messages from the original exception.

Solutions

1. Remove or set the 'catalog' field to the target database's default catalog in the uploaded YAML
2. Enable multi-catalog on the target connection if the engine supports it (update the SQLAlchemy URI / engine settings so allow_multi_catalog is true)
3. Import into a database connection whose default catalog matches the dataset's catalog

Example fix

# before (exported YAML)
catalog: hive
# after (match target default catalog)
catalog: default
# or omit the key entirely

from superset.commands.dataset.importers.v1.utils import validate_catalog

for cfg in dataset_configs:  # before import, pre-flight each dataset config
    if cfg.get('catalog'):
        db = session.query(Database).filter_by(id=cfg['database_id']).first()
        if db and (default := db.get_default_catalog()) is not None:\
           not db.allow_multi_catalog and cfg['catalog'] != default:
            cfg['catalog'] = default  # normalize to the target default
# then run the import

from superset.commands.exceptions import CommandInvalidError
try:
    client.post('/api/v1/dataset/import/', files=...)
except CommandInvalidError as ex:
    # 422-class error; ex.messages lists per-config issues — fix YAML and retry once
    ...

When it happens

Trigger: POST/PUT to /api/v1/dataset/import (or dashboard import that includes datasets) with a bundle where a dataset's config has a 'catalog' value different from the database's default catalog, and the database's engine spec has allow_multi_catalog disabled (or the engine does not support catalogs at all with a set default).

Common situations: Exporting a dataset from a multi-catalog engine (e.g. Trino, Spark with catalogs) and importing it into a deployment where the connection string pins a single catalog; feature flag or engine-spec differences between source and target instances.

Related errors

• Only the default catalog is supported for this connection
• has no valid keys
• is not a valid file
• All values must be numeric
• Data URI is not allowed.

AI-assisted analysis of apache/superset@f4587218dd (2026-08-14). Data as JSON: /api/errors/614eed0f949facc5. Report an issue: GitHub.

Appendix: source

Thrown at superset/commands/dataset/importers/v1/__init__.py:80

        # import related databases
        database_ids: dict[str, int] = {}
        for file_name, config in configs.items():
            if file_name.startswith("databases/") and config["uuid"] in database_uuids:
                database = import_database(config, overwrite=False)
                database_ids[str(database.uuid)] = database.id

        # import datasets with the correct parent ref
        for file_name, config in configs.items():
            if (
                file_name.startswith("datasets/")
                and config["database_uuid"] in database_ids
            ):
                config["database_id"] = database_ids[config["database_uuid"]]
                try:
                    import_dataset(config, overwrite=overwrite)
                except MultiCatalogDisabledValidationError as ex:
                    # surface as a 422 validation error instead of a generic 500
                    raise CommandInvalidError(
                        "; ".join(str(message) for message in ex.messages),
                        [ex],
                    ) from ex

View on GitHub (pinned to f4587218dd)