ob-metaflow 2.11.15.1__tar.gz → 2.11.15.3__tar.gz

{ob_metaflow-2.11.15.1/ob_metaflow.egg-info → ob_metaflow-2.11.15.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ob-metaflow
-Version: 2.11.15.1
+Version: 2.11.15.3
 Summary: Metaflow: More Data Science, Less Engineering
 Author: Netflix, Outerbounds & the Metaflow Community
 Author-email: help@outerbounds.co
@@ -12,7 +12,7 @@ Requires-Dist: boto3
 Requires-Dist: pylint
 Requires-Dist: kubernetes
 Provides-Extra: stubs
-Requires-Dist: ob-metaflow-stubs==2.11.15.1; extra == "stubs"
+Requires-Dist: ob-metaflow-stubs==2.11.15.3; extra == "stubs"
 
 ![Metaflow_Logo_Horizontal_FullColor_Ribbon_Dark_RGB](https://user-images.githubusercontent.com/763451/89453116-96a57e00-d713-11ea-9fa6-82b29d4d6eff.png)
 

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/metaflow_config.py

@@ -154,6 +154,13 @@ AWS_SECRETS_MANAGER_DEFAULT_REGION = from_conf("AWS_SECRETS_MANAGER_DEFAULT_REGI
 # - "projects/1234567890/secrets/foo-" -> "projects/1234567890/secrets/foo-mysecret"
 GCP_SECRET_MANAGER_PREFIX = from_conf("GCP_SECRET_MANAGER_PREFIX")
 
+# Secrets Backend - Azure Key Vault prefix. With this, users don't have to
+# specify the full https:// vault url in the @secret decorator.
+#
+# It does not make a difference if the prefix ends in a / or not. We will handle either
+# case correctly.
+AZURE_KEY_VAULT_PREFIX = from_conf("AZURE_KEY_VAULT_PREFIX")
+
 # The root directory to save artifact pulls in, when using S3 or Azure
 ARTIFACT_LOCALROOT = from_conf("ARTIFACT_LOCALROOT", os.getcwd())
 
@@ -471,6 +478,7 @@ def get_pinned_conda_libs(python_version, datastore_type):
     elif datastore_type == "azure":
         pins["azure-identity"] = ">=1.10.0"
         pins["azure-storage-blob"] = ">=12.12.0"
+        pins["azure-keyvault-secrets"] = ">=4.7.0"
     elif datastore_type == "gs":
         pins["google-cloud-storage"] = ">=2.5.0"
         pins["google-auth"] = ">=2.11.0"

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/metaflow_environment.py

@@ -124,7 +124,7 @@ class MetaflowEnvironment(object):
             cmds.append("%s -m pip install awscli boto3" % self._python())
         elif datastore_type == "azure":
             cmds.append(
-                "%s -m pip install azure-identity azure-storage-blob simple-azure-blob-downloader -qqq"
+                "%s -m pip install azure-identity azure-storage-blob azure-keyvault-secrets simple-azure-blob-downloader -qqq"
                 % self._python()
             )
         elif datastore_type == "gs":

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/plugins/__init__.py

@@ -125,6 +125,10 @@ SECRETS_PROVIDERS_DESC = [
         "gcp-secret-manager",
         ".gcp.gcp_secret_manager_secrets_provider.GcpSecretManagerSecretsProvider",
     ),
+    (
+        "az-key-vault",
+        ".azure.azure_secret_manager_secrets_provider.AzureKeyVaultSecretsProvider",
+    ),
 ]
 
 AZURE_CLIENT_PROVIDERS_DESC = [

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/plugins/airflow/airflow.py

@@ -32,6 +32,7 @@ from metaflow.metaflow_config import (
     S3_ENDPOINT_URL,
     SERVICE_HEADERS,
     SERVICE_INTERNAL_URL,
+    AZURE_KEY_VAULT_PREFIX,
 )
 
 from metaflow.metaflow_config_funcs import config_values
@@ -412,6 +413,9 @@ class Airflow(object):
         if GCP_SECRET_MANAGER_PREFIX:
             env["METAFLOW_GCP_SECRET_MANAGER_PREFIX"] = GCP_SECRET_MANAGER_PREFIX
 
+        if AZURE_KEY_VAULT_PREFIX:
+            env["METAFLOW_AZURE_KEY_VAULT_PREFIX"] = AZURE_KEY_VAULT_PREFIX
+
         env.update(additional_mf_variables)
 
         service_account = (

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/plugins/argo/argo_workflows.py

@@ -33,6 +33,7 @@ from metaflow.metaflow_config import (
     DEFAULT_METADATA,
     DEFAULT_SECRETS_BACKEND_TYPE,
     GCP_SECRET_MANAGER_PREFIX,
+    AZURE_KEY_VAULT_PREFIX,
     KUBERNETES_FETCH_EC2_METADATA,
     KUBERNETES_LABELS,
     KUBERNETES_NAMESPACE,
@@ -1420,6 +1421,7 @@ class ArgoWorkflows(object):
                 "METAFLOW_AWS_SECRETS_MANAGER_DEFAULT_REGION"
             ] = AWS_SECRETS_MANAGER_DEFAULT_REGION
             env["METAFLOW_GCP_SECRET_MANAGER_PREFIX"] = GCP_SECRET_MANAGER_PREFIX
+            env["METAFLOW_AZURE_KEY_VAULT_PREFIX"] = AZURE_KEY_VAULT_PREFIX
 
             # support for Azure
             env[

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/plugins/azure/azure_exceptions.py

@@ -10,4 +10,4 @@ class MetaflowAzureResourceError(MetaflowException):
 
 
 class MetaflowAzurePackageError(MetaflowException):
-    headline = "Missing required packages 'azure-identity' and 'azure-storage-blob'"
+    headline = "Missing required packages 'azure-identity' and 'azure-storage-blob' and 'azure-keyvault-secrets'"

ob_metaflow-2.11.15.3/metaflow/plugins/azure/azure_secret_manager_secrets_provider.py

@@ -0,0 +1,236 @@
+from metaflow.plugins.secrets import SecretsProvider
+import re
+import base64
+import codecs
+from urllib.parse import urlparse
+from metaflow.exception import MetaflowException
+import sys
+from metaflow.metaflow_config import AZURE_KEY_VAULT_PREFIX
+from metaflow.plugins.azure.azure_credential import (
+    create_cacheable_azure_credential,
+)
+
+
+class MetaflowAzureKeyVaultBadVault(MetaflowException):
+    """Raised when the secretid is fully qualified but does not have the right key vault domain"""
+
+
+class MetaflowAzureKeyVaultBadSecretType(MetaflowException):
+    """Raised when the secret type is anything except secrets"""
+
+
+class MetaflowAzureKeyVaultBadSecretPath(MetaflowException):
+    """Raised when the secret path does not match to expected length"""
+
+
+class MetaflowAzureKeyVaultBadSecretName(MetaflowException):
+    """Raised when the secret name does not match expected pattern"""
+
+
+class MetaflowAzureKeyVaultBadSecretVersion(MetaflowException):
+    """Raised when the secret version does not match expected pattern"""
+
+
+class MetaflowAzureKeyVaultBadSecret(MetaflowException):
+    """Raised when the secret does not match supported patterns in Metaflow"""
+
+
+class AzureKeyVaultSecretsProvider(SecretsProvider):
+    TYPE = "az-key-vault"
+    key_vault_domains = [
+        ".vault.azure.net",
+        ".vault.azure.cn",
+        ".vault.usgovcloudapi.net",
+        ".vault.microsoftazure.de",
+    ]
+    supported_vault_object_types = ["secrets"]
+
+    # https://learn.microsoft.com/en-us/azure/key-vault/general/about-keys-secrets-certificates has details on vault name structure
+    # Vault name and Managed HSM pool name must be a 3-24 character string, containing only 0-9, a-z, A-Z, and not consecutive -.
+    def _is_valid_vault_name(self, vault_name):
+        vault_name_pattern = r"^(?!.*--)[a-zA-Z0-9-]{3,24}$"
+        return re.match(vault_name_pattern, vault_name) is not None
+
+    # The type of the object can be, "keys", "secrets", or "certificates".
+    # Currently only secrets will be supported
+    def _is_valid_object_type(self, secret_type):
+        for type in self.supported_vault_object_types:
+            if secret_type == type:
+                return True
+        return False
+
+    # The secret name must be a 1-127 character string, starting with a letter and containing only 0-9, a-z, A-Z, and -.
+    def _is_valid_secret_name(self, secret_name):
+        secret_name_pattern = r"^[a-zA-Z][a-zA-Z0-9-]{0,126}$"
+        return re.match(secret_name_pattern, secret_name) is not None
+
+    # An object-version is a system-generated, 32 character string identifier that is optionally used to address a unique version of an object.
+    def _is_valid_object_version(self, secret_version):
+        object_version_pattern = r"^[a-zA-Z0-9]{32}$"
+        return re.match(object_version_pattern, secret_version) is not None
+
+    # This function will check if the secret_id is fully qualified url. It will return True iff the secret_id is of the form:
+    # https://myvault.vault.azure.net/secrets/mysecret/ec96f02080254f109c51a1f14cdb1931 OR
+    # https://myvault.vault.azure.net/secrets/mysecret/
+    # validating the above as per recommendations in https://devblogs.microsoft.com/azure-sdk/guidance-for-applications-using-the-key-vault-libraries/
+    def _is_secret_id_fully_qualified_url(self, secret_id):
+        # if the secret_id is None/empty/does not start with https then return false
+        if secret_id is None or secret_id == "" or not secret_id.startswith("https://"):
+            return False
+        try:
+            parsed_vault_url = urlparse(secret_id)
+        except ValueError:
+            print(f"invalid vault url", file=sys.stderr)
+            return False
+        hostname = parsed_vault_url.netloc
+
+        k_v_domain_found = False
+        actual_k_v_domain = ""
+        for k_v_domain in self.key_vault_domains:
+            if k_v_domain in hostname:
+                k_v_domain_found = True
+                actual_k_v_domain = k_v_domain
+                break
+        if not k_v_domain_found:
+            # the secret_id started with https:// however the key_vault_domains
+            # were not present in the secret_id which means
+            raise MetaflowAzureKeyVaultBadVault(f"bad key vault domain {secret_id}")
+
+        # given the secret_id seems to have a valid key vault domain
+        # lets verify that the vault name corresponds to its regex.
+        vault_name = hostname[: -len(actual_k_v_domain)]
+        # verify the vault name pattern
+        if not self._is_valid_vault_name(vault_name):
+            raise MetaflowAzureKeyVaultBadVault(f"bad key vault name {vault_name}")
+
+        path_parts = parsed_vault_url.path.strip("/").split("/")
+        total_path_parts = len(path_parts)
+        if total_path_parts < 2 or total_path_parts > 3:
+            raise MetaflowAzureKeyVaultBadSecretPath(
+                f"bad secret uri path {path_parts}"
+            )
+
+        object_type = path_parts[0]
+        if not self._is_valid_object_type(object_type):
+            raise MetaflowAzureKeyVaultBadSecretType(f"bad secret type {object_type}")
+
+        secret_name = path_parts[1]
+        if not self._is_valid_secret_name(secret_name=secret_name):
+            raise MetaflowAzureKeyVaultBadSecretName(f"bad secret name {secret_name}")
+
+        if total_path_parts == 3:
+            if not self._is_valid_object_version(path_parts[2]):
+                raise MetaflowAzureKeyVaultBadSecretVersion(
+                    f"bad secret version {path_parts[2]}"
+                )
+
+        return True
+
+    # This function will validate the correctness of the partial secret id.
+    # It will attempt to construct the fully qualified secret URL internally and
+    # call the _is_secret_id_fully_qualified_url to check validity
+    def _is_partial_secret_valid(self, secret_id):
+        secret_parts = secret_id.strip("/").split("/")
+        total_secret_parts = len(secret_parts)
+        if total_secret_parts < 1 or total_secret_parts > 2:
+            return False
+
+        # since the secret_id is supposedly a partial id, the AZURE_KEY_VAULT_PREFIX
+        # must be set.
+        if not AZURE_KEY_VAULT_PREFIX:
+            raise ValueError(
+                f"cannot use simple secret id without setting METAFLOW_AZURE_KEY_VAULT_PREFIX. {AZURE_KEY_VAULT_PREFIX}"
+            )
+        domain = AZURE_KEY_VAULT_PREFIX.rstrip("/")
+        full_secret = f"{domain}/secrets/{secret_id}"
+        if not self._is_secret_id_fully_qualified_url(full_secret):
+            return False
+
+        return True
+
+    def _sanitize_key_as_env_var(self, key):
+        """
+        Sanitize a key as an environment variable name.
+        This is purely a convenience trade-off to cover common cases well, vs. introducing
+        ambiguities (e.g. did the final '_' come from '.', or '-' or is original?).
+
+        1/27/2023(jackie):
+
+        We start with few rules and should *sparingly* add more over time.
+        Also, it's TBD whether all possible providers will share the same sanitization logic.
+        Therefore we will keep this function private for now
+        """
+        return key.replace("-", "_").replace(".", "_").replace("/", "_")
+
+    def get_secret_as_dict(self, secret_id, options={}, role=None):
+        # https://learn.microsoft.com/en-us/azure/app-service/app-service-key-vault-references?tabs=azure-cli has a lot of details on
+        # the patterns used in key vault
+        # Vault names and Managed HSM pool names are selected by the user and are globally unique.
+        # Vault name and Managed HSM pool name must be a 3-24 character string, containing only 0-9, a-z, A-Z, and not consecutive -.
+        # object-type	The type of the object. As of 05/08/24 only "secrets", are supported
+        # object-name	An object-name is a user provided name for and must be unique within a key vault. The name must be a 1-127 character string, starting with a letter and containing only 0-9, a-z, A-Z, and -.
+        # object-version	An object-version is a system-generated, 32 character string identifier that is optionally used to address a unique version of an object.
+
+        # We allow these forms of secret_id:
+        #
+        # 1. Full path like https://<key-vault-name><.vault-domain>/secrets/<secret-name>/<secret-version>. This is what you
+        # see in Azure portal and is easy to copy paste.
+        #
+        # 2. Full path but without the version like https://<key-vault-name><.vault-domain>/secrets/<secret-name>
+        #
+        # 3. Simple string like mysecret. This corresponds to the SecretName.
+        #
+        # 4. Simple string with <secret-name>/<secret-version> suffix like mysecret/123
+
+        # The latter two forms require METAFLOW_AZURE_KEY_VAULT_PREFIX to be set.
+
+        # if the secret_id is None/empty/does not start with https then return false
+        if secret_id is None or secret_id == "":
+            raise MetaflowAzureKeyVaultBadSecret(f"empty secret id is not supported")
+
+        # check if the passed in secret is a short-form ( #3/#4 in the above comment)
+        if not secret_id.startswith("https://"):
+            # check if the secret_id is of form `secret_name` OR `secret_name/secret_version`
+            if not self._is_partial_secret_valid(secret_id=secret_id):
+                raise MetaflowAzureKeyVaultBadSecret(
+                    f"unsupported partial secret {secret_id}"
+                )
+
+            domain = AZURE_KEY_VAULT_PREFIX.rstrip("/")
+            full_secret = f"{domain}/secrets/{secret_id}"
+
+        # if the secret id is passed as a URL - then check if the url is fully qualified
+        if secret_id.startswith("https://"):
+            if not self._is_secret_id_fully_qualified_url(secret_id=secret_id):
+                raise MetaflowException(f"unsupported secret {secret_id}")
+            full_secret = secret_id
+
+        # at this point I know that the secret URL is good so we can start creating the Secret Client
+        az_credentials = create_cacheable_azure_credential()
+        res = urlparse(full_secret)
+        az_vault_url = f"{res.scheme}://{res.netloc}"  # https://myvault.vault.azure.net
+        secret_data = res.path.strip("/").split("/")[1:]
+        secret_name = secret_data[0]
+        secret_version = None
+        if len(secret_data) > 1:
+            secret_version = secret_data[1]
+
+        from azure.keyvault.secrets import SecretClient
+
+        client = SecretClient(vault_url=az_vault_url, credential=az_credentials)
+
+        key_vault_secret_val = client.get_secret(
+            name=secret_name, version=secret_version
+        )
+
+        result = {}
+
+        if options.get("env_var_name") is not None:
+            env_var_name = options["env_var_name"]
+            sanitized_key = self._sanitize_key_as_env_var(env_var_name)
+        else:
+            sanitized_key = self._sanitize_key_as_env_var(key_vault_secret_val.name)
+
+        response_payload = key_vault_secret_val.value
+        result[sanitized_key] = response_payload
+        return result

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/metaflow/plugins/kubernetes/kubernetes.py

@@ -32,6 +32,7 @@ from metaflow.metaflow_config import (
     DEFAULT_METADATA,
     DEFAULT_SECRETS_BACKEND_TYPE,
     GCP_SECRET_MANAGER_PREFIX,
+    AZURE_KEY_VAULT_PREFIX,
     KUBERNETES_FETCH_EC2_METADATA,
     KUBERNETES_LABELS,
     KUBERNETES_SANDBOX_INIT_SCRIPT,
@@ -261,6 +262,9 @@ class Kubernetes(object):
             .environment_variable(
                 "METAFLOW_GCP_SECRET_MANAGER_PREFIX", GCP_SECRET_MANAGER_PREFIX
             )
+            .environment_variable(
+                "METAFLOW_AZURE_KEY_VAULT_PREFIX", AZURE_KEY_VAULT_PREFIX
+            )
             .environment_variable("METAFLOW_S3_ENDPOINT_URL", S3_ENDPOINT_URL)
             .environment_variable(
                 "METAFLOW_AZURE_STORAGE_BLOB_SERVICE_ENDPOINT",

ob_metaflow-2.11.15.3/metaflow/version.py

@@ -0,0 +1 @@
+metaflow_version = "2.11.15.3"

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3/ob_metaflow.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ob-metaflow
-Version: 2.11.15.1
+Version: 2.11.15.3
 Summary: Metaflow: More Data Science, Less Engineering
 Author: Netflix, Outerbounds & the Metaflow Community
 Author-email: help@outerbounds.co
@@ -12,7 +12,7 @@ Requires-Dist: boto3
 Requires-Dist: pylint
 Requires-Dist: kubernetes
 Provides-Extra: stubs
-Requires-Dist: ob-metaflow-stubs==2.11.15.1; extra == "stubs"
+Requires-Dist: ob-metaflow-stubs==2.11.15.3; extra == "stubs"
 
 ![Metaflow_Logo_Horizontal_FullColor_Ribbon_Dark_RGB](https://user-images.githubusercontent.com/763451/89453116-96a57e00-d713-11ea-9fa6-82b29d4d6eff.png)
 

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/ob_metaflow.egg-info/SOURCES.txt

@@ -182,6 +182,7 @@ metaflow/plugins/aws/step_functions/step_functions_decorator.py
 metaflow/plugins/azure/__init__.py
 metaflow/plugins/azure/azure_credential.py
 metaflow/plugins/azure/azure_exceptions.py
+metaflow/plugins/azure/azure_secret_manager_secrets_provider.py
 metaflow/plugins/azure/azure_tail.py
 metaflow/plugins/azure/azure_utils.py
 metaflow/plugins/azure/blob_service_client_factory.py

{ob_metaflow-2.11.15.1 → ob_metaflow-2.11.15.3}/ob_metaflow.egg-info/requires.txt

@@ -4,4 +4,4 @@ pylint
 kubernetes
 
 [stubs]
-ob-metaflow-stubs==2.11.15.1
+ob-metaflow-stubs==2.11.15.3

ob_metaflow-2.11.15.1/metaflow/version.py

@@ -1 +0,0 @@
-metaflow_version = "2.11.15.1"