azure-storage-blob-py
用 Python 對 Azure Blob Storage 上傳、下載、列出與管理 blob
中文翻譯
---
name: azure-storage-blob-py
description: |
Python 版的 Azure Blob Storage SDK。用於上傳、下載、列出 blob、管理 container 和 blob 生命週期。
Triggers: "blob storage", "BlobServiceClient", "ContainerClient", "BlobClient", "upload blob", "download blob".
license: MIT
metadata:
author: Microsoft
version: "1.0.0"
package: azure-storage-blob
---
# Python 版 Azure Blob Storage SDK
Azure Blob Storage 的 client library，Blob Storage 是用來存放非結構化資料的物件儲存。
## 安裝
```bash
pip install azure-storage-blob azure-identity
```
## 環境變數
```bash
AZURE_STORAGE_ACCOUNT_NAME=<your-storage-account> # 所有驗證方式都必填
# 或用完整 URL
AZURE_STORAGE_ACCOUNT_URL=https://<account>.blob.core.windows.net # account name 的替代方案
AZURE_TOKEN_CREDENTIALS=prod # 只有在 production 使用 DefaultAzureCredential 時才需要
```
## 驗證與生命週期
> **🔑 下面每個程式碼範例都適用兩條規則：**
>
> 1. **優先用 `DefaultAzureCredential`。** 它在本機（Azure CLI / VS Code / Developer CLI）和 Azure 上（managed identity、workload identity）都能用，不用改程式碼。避免用 connection string、account key 或 API key，它們會繞過 Entra 的稽核和輪替。
> - 本機開發：直接用 `DefaultAzureCredential` 就好。
> - Production：設定 `AZURE_TOKEN_CREDENTIALS=prod`（或 `AZURE_TOKEN_CREDENTIALS=<specific_credential>`），把 credential chain 限縮在 production 安全的 credential。
> 2. **每個 client 都用 context manager 包起來**，這樣 HTTP transport、socket 和 token cache 才會確實釋放：
> - 同步：`with <Client>(...) as client:`
> - 非同步：`async with <Client>(...) as client:` **而且** `async with DefaultAzureCredential() as credential:`（來自 `azure.identity.aio`）
>
> 範例片段可能會簡化這段設定，但 production 的程式碼一定要兩條規則都遵守。
```python
from azure.identity import DefaultAzureCredential, ManagedIdentityCredential
from azure.storage.blob import BlobServiceClient
# 本機開發：DefaultAzureCredential。Production：設定 AZURE_TOKEN_CREDENTIALS=prod 或 AZURE_TOKEN_CREDENTIALS=<specific_credential>
credential = DefaultAzureCredential(require_envvar=True)
# 或在 production 直接用特定的 credential：
# 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:
# 在這裡使用 blob_service_client（各種操作請看後面的章節）
...
```
## Client 階層
| Client | 用途 | 取得方式 |
|--------|---------|----------|
| `BlobServiceClient` | Account 層級的操作 | 直接建立實例 |
| `ContainerClient` | Container 操作 | `blob_service_client.get_container_client()` |
| `BlobClient` | 單一 blob 的操作 | `container_client.get_blob_client()` |
## 核心工作流程
### 建立 Container
```python
container_client = blob_service_client.get_container_client("mycontainer")
container_client.create_container()
```
### 上傳 Blob
```python
# 從檔案路徑
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)
# 從 bytes/字串
blob_client.upload_blob(b"Hello, World!", overwrite=True)
# 從 stream
import io
stream = io.BytesIO(b"Stream content")
blob_client.upload_blob(stream, overwrite=True)
```
### 下載 Blob
```python
blob_client = blob_service_client.get_blob_client(
container="mycontainer",
blob="sample.txt"
)
# 下載到檔案
with open("./downloaded.txt", "wb") as file:
download_stream = blob_client.download_blob()
file.write(download_stream.readall())
# 下載到記憶體
download_stream = blob_client.download_blob()
content = download_stream.readall() # bytes
# 讀進既有的 buffer
stream = io.BytesIO()
num_bytes = blob_client.download_blob().readinto(stream)
```
### 列出 Blob
```python
container_client = blob_service_client.get_container_client("mycontainer")
# 列出所有 blob
for blob in container_client.list_blobs():
print(f"{blob.name} - {blob.size} bytes")
# 用 prefix 列出（類似資料夾）
for blob in container_client.list_blobs(name_starts_with="logs/"):
print(blob.name)
# 走訪 blob 階層（虛擬目錄）
for item in container_client.walk_blobs(delimiter="/"):
if item.get("prefix"):
print(f"Directory: {item['prefix']}")
else:
print(f"Blob: {item.name}")
```
### 刪除 Blob
```python
blob_client.delete_blob()
# 連同 snapshot 一起刪除
blob_client.delete_blob(delete_snapshots="include")
```
## 效能調校
```python
# 為大型上傳／下載設定 chunk 大小
with BlobClient(
account_url=account_url,
container_name="mycontainer",
blob_name="large-file.zip",
credential=credential,
max_block_size=4 * 1024 * 1024, # 4 MiB 的 block
max_single_put_size=64 * 1024 * 1024 # 單次上傳上限 64 MiB
) as blob_client:
# 平行上傳
blob_client.upload_blob(data, max_concurrency=4)
# 平行下載
download_stream = blob_client.download_blob(max_concurrency=4)
```
## SAS Token（User Delegation）
SAS token 要用 Microsoft Entra ID 簽署的 **user delegation key** 來產生，絕對不要用 account key。這樣 SAS 的核發才會綁在 Entra 的稽核和輪替機制上。
```python
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:
# 取得 user delegation key（最長有效 7 天）。呼叫者需要在 storage account 上
# 擁有 "Storage Blob Delegator" 角色。
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 屬性與 Metadata
```python
# 取得屬性
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}")
# 設定 metadata
blob_client.set_blob_metadata(metadata={"category": "logs", "year": "2024"})
# 設定 content type
from azure.storage.blob import ContentSettings
blob_client.set_http_headers(
content_settings=ContentSettings(content_type="application/json")
)
```
## 非同步 Client
```python
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)
# 非同步下載
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()
```
## 最佳實務
1. **同步或非同步選一個，然後保持一致。** 同一條呼叫路徑裡，不要混用 `azure.storage.blob` 同步 client 和 `azure.storage.blob.aio` 非同步 client。每個 module 只選一種模式。
2. **client 和非同步 credential 一律用 context manager。** 每個 client 都用 `with BlobServiceClient(...) as client:`（同步）或 `async with BlobServiceClient(...) as client:`（非同步）包起來。非同步的 `DefaultAzureCredential`（來自 `azure.identity.aio`）也要用 `async with credential:`，token 和 transport 才會清乾淨。
3. 在本機跑的程式碼**用 `DefaultAzureCredential`**（不要用 connection string）。在 Azure 上跑的程式碼用特定的 token credential。
4. 重新上傳時，**明確設定 `overwrite=True`**
5. 大檔案傳輸**用 `max_concurrency`**
6. **優先用 `readinto()`** 而不是 `readall()`，比較省記憶體
7. 階層式列出**用 `walk_blobs()`**
8. 給網頁提供的 blob **設定適當的 content type**
## 參考檔案
| 檔案 | 內容 |
|------|----------|
| [references/capabilities.md](references/capabilities.md) | 能力索引，對應主要流程和非主要流程的參考文件。 |
| [references/non-hero-scenarios.md](references/non-hero-scenarios.md) | 專屬的非主要流程範例（metadata/屬性和非同步模式）。 |英文原文
---
name: azure-storage-blob-py
description: |
Azure Blob Storage SDK for Python. Use for uploading, downloading, listing blobs, managing containers, and blob lifecycle.
Triggers: "blob storage", "BlobServiceClient", "ContainerClient", "BlobClient", "upload blob", "download blob".
license: MIT
metadata:
author: Microsoft
version: "1.0.0"
package: azure-storage-blob
---
# Azure Blob Storage SDK for Python
Client library for Azure Blob Storage — object storage for unstructured data.
## Installation
```bash
pip install azure-storage-blob azure-identity
```
## Environment Variables
```bash
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.
```python
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
```python
container_client = blob_service_client.get_container_client("mycontainer")
container_client.create_container()
```
### Upload Blob
```python
# 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
```python
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
```python
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
```python
blob_client.delete_blob()
# Delete with snapshots
blob_client.delete_blob(delete_snapshots="include")
```
## Performance Tuning
```python
# 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.
```python
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
```python
# 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
```python
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](references/capabilities.md) | Capability index mapping hero flows and non-hero references. |
| [references/non-hero-scenarios.md](references/non-hero-scenarios.md) | Dedicated non-hero examples (metadata/properties and async patterns). |