def _upload_artifacts_to_databricks(
source, run_id, source_host_uri=None, target_databricks_profile_uri=None
):
"""
Copy the artifacts from ``source`` to the destination Databricks workspace (DBFS) given by
``databricks_profile_uri`` or the current tracking URI.
:param source: Source location for the artifacts to copy.
:param run_id: Run ID to associate the artifacts with.
:param source_host_uri: Specifies the source artifact's host URI (e.g. Databricks tracking URI)
if applicable. If not given, defaults to the current tracking URI.
:param target_databricks_profile_uri: Specifies the destination Databricks host. If not given,
defaults to the current tracking URI.
:return: The DBFS location in the target Databricks workspace the model files have been
uploaded to.
"""
from uuid import uuid4
local_dir = tempfile.mkdtemp()
try:
source_with_profile = add_databricks_profile_info_to_artifact_uri(source, source_host_uri)
_download_artifact_from_uri(source_with_profile, local_dir)
dest_root = "dbfs:/databricks/mlflow/tmp-external-source/"
dest_root_with_profile = add_databricks_profile_info_to_artifact_uri(
dest_root, target_databricks_profile_uri
)
dest_repo = DbfsRestArtifactRepository(dest_root_with_profile)
dest_artifact_path = run_id if run_id else uuid4().hex
# Allow uploading from the same run id multiple times by randomizing a suffix
if len(dest_repo.list_artifacts(dest_artifact_path)) > 0:
dest_artifact_path = dest_artifact_path + "-" + uuid4().hex[0:4]
dest_repo.log_artifacts(local_dir, artifact_path=dest_artifact_path)
dirname = pathlib.PurePath(source).name # innermost directory name
return posixpath.join(dest_root, dest_artifact_path, dirname) # new source
finally:
shutil.rmtree(local_dir)
def test_init_validation_and_cleaning(self):
with mock.patch(DBFS_ARTIFACT_REPOSITORY_PACKAGE + '._get_host_creds_from_default_store') \
as get_creds_mock:
get_creds_mock.return_value = lambda: MlflowHostCreds('http://host'
)
repo = get_artifact_repository('dbfs:/test/')
assert repo.artifact_uri == 'dbfs:/test'
with pytest.raises(MlflowException):
DbfsRestArtifactRepository('s3://test')
with pytest.raises(MlflowException):
DbfsRestArtifactRepository(
'dbfs://profile@notdatabricks/test/')
def test_init_validation_and_cleaning(self):
with mock.patch(
DBFS_ARTIFACT_REPOSITORY_PACKAGE +
"._get_host_creds_from_default_store") as get_creds_mock:
get_creds_mock.return_value = lambda: MlflowHostCreds("http://host"
)
repo = get_artifact_repository("dbfs:/test/")
assert repo.artifact_uri == "dbfs:/test"
match = "DBFS URI must be of the form dbfs:/<path>"
with pytest.raises(MlflowException, match=match):
DbfsRestArtifactRepository("s3://test")
with pytest.raises(MlflowException, match=match):
DbfsRestArtifactRepository(
"dbfs://profile@notdatabricks/test/")
def get_dbfs_artifact_repo(self, artifact_uri):
try:
from mlflow.store.artifact.dbfs_artifact_repo import DbfsRestArtifactRepository
except ImportError:
logger.warning(
VERSION_WARNING.format(
"DbfsRestArtifactRepository from " +
"mlflow.store.artifact.dbfs_artifact_repo"))
from mlflow.store.dbfs_artifact_repo import DbfsRestArtifactRepository
# dbfs artifact_uri has to have dbfs as prefix
# by default it is dbfs:/databricks/mlflow/<exp_id>/<run_id>/artifacts
artifact_uri = artifact_uri.replace("adbazureml", "dbfs", 1)
logger.info("DBFS artifact uri is {}".format(artifact_uri))
return DbfsRestArtifactRepository(artifact_uri)
def test_init_get_host_creds_with_databricks_profile_uri(self):
databricks_host = 'https://something.databricks.com'
default_host = 'http://host'
with mock.patch(DBFS_ARTIFACT_REPOSITORY_PACKAGE + '._get_host_creds_from_default_store',
return_value=lambda: MlflowHostCreds(default_host)), \
mock.patch(DBFS_ARTIFACT_REPOSITORY_PACKAGE + '.get_databricks_host_creds',
return_value=MlflowHostCreds(databricks_host)):
repo = DbfsRestArtifactRepository(
'dbfs://profile@databricks/test/')
assert repo.artifact_uri == 'dbfs:/test/'
creds = repo.get_host_creds()
assert creds.host == databricks_host
# no databricks_profile_uri given
repo = DbfsRestArtifactRepository('dbfs:/test/')
creds = repo.get_host_creds()
assert creds.host == default_host
def test_init_get_host_creds_with_databricks_profile_uri(self):
databricks_host = "https://something.databricks.com"
default_host = "http://host"
with mock.patch(
DBFS_ARTIFACT_REPOSITORY_PACKAGE +
"._get_host_creds_from_default_store",
return_value=lambda: MlflowHostCreds(default_host),
), mock.patch(
DBFS_ARTIFACT_REPOSITORY_PACKAGE +
".get_databricks_host_creds",
return_value=MlflowHostCreds(databricks_host),
):
repo = DbfsRestArtifactRepository(
"dbfs://profile@databricks/test/")
assert repo.artifact_uri == "dbfs:/test/"
creds = repo.get_host_creds()
assert creds.host == databricks_host
# no databricks_profile_uri given
repo = DbfsRestArtifactRepository("dbfs:/test/")
creds = repo.get_host_creds()
assert creds.host == default_host
def dbfs_artifact_repo_factory(artifact_uri):
"""
Returns an ArtifactRepository subclass for storing artifacts on DBFS.
This factory method is used with URIs of the form ``dbfs:/<path>``. DBFS-backed artifact
storage can only be used together with the RestStore.
In the special case where the URI is of the form
`dbfs:/databricks/mlflow-tracking/<Exp-ID>/<Run-ID>/<path>',
a DatabricksArtifactRepository is returned. This is capable of storing access controlled
artifacts.
:param artifact_uri: DBFS root artifact URI (string).
:return: Subclass of ArtifactRepository capable of storing artifacts on DBFS.
"""
try:
if supports_acled_artifacts(mlflow.__version__):
from mlflow.store.artifact.dbfs_artifact_repo import dbfs_artifact_repo_factory
return dbfs_artifact_repo_factory(artifact_uri)
except Exception:
pass
# For some reason, we must import modules specific to this package within the
# entrypoint function rather than the top-level module. Otherwise, entrypoint
# registration fails with import errors
from mlflow_databricks_artifacts.store.artifact_repo import DatabricksArtifactRepository
from mlflow_databricks_artifacts.utils.databricks_utils import is_dbfs_fuse_available
from mlflow_databricks_artifacts.utils.string_utils import strip_prefix
from mlflow_databricks_artifacts.utils.uri import (
get_databricks_profile_uri_from_artifact_uri,
is_databricks_acled_artifacts_uri,
is_databricks_model_registry_artifacts_uri,
is_valid_dbfs_uri,
remove_databricks_profile_info_from_artifact_uri,
)
if not is_valid_dbfs_uri(artifact_uri):
raise MlflowException(
"DBFS URI must be of the form dbfs:/<path> or "
+ "dbfs://profile@databricks/<path>, but received "
+ artifact_uri
)
cleaned_artifact_uri = artifact_uri.rstrip("/")
db_profile_uri = get_databricks_profile_uri_from_artifact_uri(cleaned_artifact_uri)
if is_databricks_acled_artifacts_uri(artifact_uri):
return DatabricksArtifactRepository(cleaned_artifact_uri)
elif (
is_dbfs_fuse_available()
and os.environ.get(USE_FUSE_ENV_VAR, "").lower() != "false"
and not is_databricks_model_registry_artifacts_uri(artifact_uri)
and (db_profile_uri is None or db_profile_uri == "databricks")
):
# If the DBFS FUSE mount is available, write artifacts directly to
# /dbfs/... using local filesystem APIs.
# Note: it is possible for a named Databricks profile to point to the current workspace,
# but we're going to avoid doing a complex check and assume users will use `databricks`
# to mean the current workspace. Using `DbfsRestArtifactRepository` to access the current
# workspace's DBFS should still work; it just may be slower.
final_artifact_uri = remove_databricks_profile_info_from_artifact_uri(cleaned_artifact_uri)
file_uri = "file:///dbfs/{}".format(strip_prefix(final_artifact_uri, "dbfs:/"))
return LocalArtifactRepository(file_uri)
return DbfsRestArtifactRepository(cleaned_artifact_uri)
