Azure Blob Storage SDK for Python

Client library for Azure Blob Storage — object storage for unstructured data.

Installation

pip install azure-storage-blob azure-identity

Environment Variables

AZURE_STORAGE_ACCOUNT_NAME=<your-storage-account>  # Required for all auth methods
# Or use full URL
AZURE_STORAGE_ACCOUNT_URL=https://<account>.blob.core.windows.net  # Alternative to account name
AZURE_TOKEN_CREDENTIALS=prod # Required only if DefaultAzureCredential is used in production

Authentication & Lifecycle

🔑 Two rules apply to every code sample below:

1. Prefer DefaultAzureCredential. It works locally (Azure CLI / VS Code / Developer CLI) and in Azure (managed identity, workload identity) with no code change. Avoid connection strings, account/API keys — they bypass Entra audit and rotation.
   • Local dev: DefaultAzureCredential works as-is.
   • Production: set AZURE_TOKEN_CREDENTIALS=prod (or AZURE_TOKEN_CREDENTIALS=<specific_credential>) to constrain the credential chain to production-safe credentials.
2. Wrap every client in a context manager so HTTP transports, sockets, and token caches are released deterministically:
   • Sync: with <Client>(...) as client:
   • Async: async with <Client>(...) as client: and async with DefaultAzureCredential() as credential: (from azure.identity.aio)

Snippets may abbreviate this setup, but production code should always follow both rules.

from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.storage.blob import BlobServiceClient

# Local dev: DefaultAzureCredential. Production: set AZURE_TOKEN_CREDENTIALS=prod or AZURE_TOKEN_CREDENTIALS=<specific_credential>
credential = DefaultAzureCredential(require_envvar=True)
# Or use a specific credential directly in production:
# See https://learn.microsoft.com/python/api/overview/azure/identity-readme?view=azure-python#credential-classes
# credential = ManagedIdentityCredential()
account_url = "https://<account>.blob.core.windows.net"

with BlobServiceClient(account_url, credential=credential) as blob_service_client:
    # Use blob_service_client here (see following sections for operations)
    ...

Client Hierarchy

Client	Purpose	Get From
BlobServiceClient	Account-level operations	Direct instantiation
ContainerClient	Container operations	blob_service_client.get_container_client()
BlobClient	Single blob operations	container_client.get_blob_client()

Core Workflow

Create Container

container_client = blob_service_client.get_container_client("mycontainer")
container_client.create_container()

Upload Blob

# From file path
blob_client = blob_service_client.get_blob_client(
    container="mycontainer",
    blob="sample.txt"
)

with open("./local-file.txt", "rb") as data:
    blob_client.upload_blob(data, overwrite=True)

# From bytes/string
blob_client.upload_blob(b"Hello, World!", overwrite=True)

# From stream
import io
stream = io.BytesIO(b"Stream content")
blob_client.upload_blob(stream, overwrite=True)

Download Blob

blob_client = blob_service_client.get_blob_client(
    container="mycontainer",
    blob="sample.txt"
)

# To file
with open("./downloaded.txt", "wb") as file:
    download_stream = blob_client.download_blob()
    file.write(download_stream.readall())

# To memory
download_stream = blob_client.download_blob()
content = download_stream.readall()  # bytes

# Read into existing buffer
stream = io.BytesIO()
num_bytes = blob_client.download_blob().readinto(stream)

List Blobs

container_client = blob_service_client.get_container_client("mycontainer")

# List all blobs
for blob in container_client.list_blobs():
    print(f"{blob.name} - {blob.size} bytes")

# List with prefix (folder-like)
for blob in container_client.list_blobs(name_starts_with="logs/"):
    print(blob.name)

# Walk blob hierarchy (virtual directories)
for item in container_client.walk_blobs(delimiter="/"):
    if item.get("prefix"):
        print(f"Directory: {item['prefix']}")
    else:
        print(f"Blob: {item.name}")

Delete Blob

blob_client.delete_blob()

# Delete with snapshots
blob_client.delete_blob(delete_snapshots="include")

Performance Tuning

# Configure chunk sizes for large uploads/downloads
with BlobClient(
    account_url=account_url,
    container_name="mycontainer",
    blob_name="large-file.zip",
    credential=credential,
    max_block_size=4 * 1024 * 1024,  # 4 MiB blocks
    max_single_put_size=64 * 1024 * 1024  # 64 MiB single upload limit
) as blob_client:
    # Parallel upload
    blob_client.upload_blob(data, max_concurrency=4)

    # Parallel download
    download_stream = blob_client.download_blob(max_concurrency=4)

SAS Tokens (User Delegation)

Generate SAS tokens with a user delegation key signed by Microsoft Entra ID — never with an account key. This keeps SAS issuance tied to Entra audit/rotation.

from datetime import datetime, timedelta, timezone
from azure.identity import DefaultAzureCredential
from azure.storage.blob import (
    BlobServiceClient,
    BlobSasPermissions,
    generate_blob_sas,
)

now = datetime.now(timezone.utc)
account_url = "https://<account>.blob.core.windows.net"

with BlobServiceClient(account_url, credential=DefaultAzureCredential()) as service:
    # Get a user delegation key (valid up to 7 days). Caller needs the
    # "Storage Blob Delegator" role on the storage account.
    udk = service.get_user_delegation_key(
        key_start_time=now,
        key_expiry_time=now + timedelta(hours=1),
    )

    sas_token = generate_blob_sas(
        account_name="<account>",
        container_name="mycontainer",
        blob_name="sample.txt",
        user_delegation_key=udk,
        permission=BlobSasPermissions(read=True),
        expiry=now + timedelta(hours=1),
    )

blob_url = f"{account_url}/mycontainer/sample.txt?{sas_token}"

Blob Properties and Metadata

# Get properties
properties = blob_client.get_blob_properties()
print(f"Size: {properties.size}")
print(f"Content-Type: {properties.content_settings.content_type}")
print(f"Last modified: {properties.last_modified}")

# Set metadata
blob_client.set_blob_metadata(metadata={"category": "logs", "year": "2024"})

# Set content type
from azure.storage.blob import ContentSettings
blob_client.set_http_headers(
    content_settings=ContentSettings(content_type="application/json")
)

Async Client

from azure.identity.aio import DefaultAzureCredential
from azure.storage.blob.aio import BlobServiceClient

async def upload_async():
    async with DefaultAzureCredential() as credential:
        async with BlobServiceClient(account_url, credential=credential) as client:
            blob_client = client.get_blob_client("mycontainer", "sample.txt")
            
            with open("./file.txt", "rb") as data:
                await blob_client.upload_blob(data, overwrite=True)

# Download async
async def download_async():
    async with BlobServiceClient(account_url, credential=credential) as client:
        blob_client = client.get_blob_client("mycontainer", "sample.txt")
        
        stream = await blob_client.download_blob()
        data = await stream.readall()

Best Practices

1. Pick sync OR async and stay consistent. Do not mix azure.storage.blob sync clients with azure.storage.blob.aio async clients in the same call path. Choose one mode per module.
2. Always use context managers for clients and async credentials. Wrap every client in with BlobServiceClient(...) as client: (sync) or async with BlobServiceClient(...) as client: (async). For async DefaultAzureCredential from azure.identity.aio, also use async with credential: so tokens and transports are cleaned up.
3. Use DefaultAzureCredential for code that runs locally (instead of connection strings). Use a specific token credential for code that runs in Azure.
4. Set overwrite=True explicitly when re-uploading
5. Use max_concurrency for large file transfers
6. Prefer readinto() over readall() for memory efficiency
7. Use walk_blobs() for hierarchical listing
8. Set appropriate content types for web-served blobs

Reference Files

File	Contents
references/capabilities.md	Capability index mapping hero flows and non-hero references.
references/non-hero-scenarios.md	Dedicated non-hero examples (metadata/properties and async patterns).