Fix S3 storage to honor region configuration

example_config.yml

@@ -125,6 +125,9 @@ storages:
       aws_access_key_id: some_key
       aws_secret_access_key: some_secret
       aws_bucket: some_bucket
+      aws_region: us-east-1  # AWS region where the bucket is located
+      # endpoint_url: null  # Optional: custom endpoint for S3-compatible services
+      # s3_path: vcons      # Optional: prefix for S3 keys
   milvus:
     module: storage.milvus
     options:

server/storage/s3/README.md

@@ -8,31 +8,51 @@ S3 storage provides scalable, durable object storage capabilities, making it ide
 
 ## Configuration
 
-Required configuration options:
+Configuration options:
 
 ```yaml
 storages:
   s3:
     module: storage.s3
     options:
-      bucket: your-bucket-name           # S3 bucket name
-      region: us-west-2                  # AWS region
-      access_key: your-access-key        # AWS access key
-      secret_key: your-secret-key        # AWS secret key
-      prefix: vcons/                     # Optional: key prefix
-      endpoint_url: null                 # Optional: custom endpoint
+      # Required options
+      aws_access_key_id: your-access-key      # AWS access key ID
+      aws_secret_access_key: your-secret-key  # AWS secret access key
+      aws_bucket: your-bucket-name            # S3 bucket name
+
+      # Optional options
+      aws_region: us-east-1                   # AWS region (recommended to avoid cross-region errors)
+      endpoint_url: null                      # Custom endpoint for S3-compatible services (e.g., MinIO)
+      s3_path: vcons                          # Prefix for S3 keys (optional)
+```
+
+### Configuration Options
+
+| Option | Required | Description |
+|--------|----------|-------------|
+| `aws_access_key_id` | Yes | AWS access key ID for authentication |
+| `aws_secret_access_key` | Yes | AWS secret access key for authentication |
+| `aws_bucket` | Yes | Name of the S3 bucket to store vCons |
+| `aws_region` | No | AWS region where the bucket is located (e.g., `us-east-1`, `us-west-2`, `eu-west-1`). **Recommended** to avoid "AuthorizationHeaderMalformed" errors when the bucket is in a different region than the default. |
+| `endpoint_url` | No | Custom endpoint URL for S3-compatible services like MinIO, LocalStack, or other providers |
+| `s3_path` | No | Prefix path for organizing vCon objects within the bucket |
+
+### Region Configuration
+
+**Important:** If your S3 bucket is in a region other than `us-east-1`, you should explicitly set the `aws_region` option. Without this, you may encounter errors like:
+
+```
+AuthorizationHeaderMalformed: The authorization header is malformed;
+the region 'us-east-1' is wrong; expecting 'us-east-2'
 ```
 
 ## Features
 
-- Object storage
-- High availability
-- Durability
-- Versioning support
-- Lifecycle management
-- Automatic metrics logging
-- Encryption support
-- Access control
+- Object storage with automatic date-based key organization (`YYYY/MM/DD/uuid.vcon`)
+- High availability and durability
+- Support for custom S3-compatible endpoints (MinIO, LocalStack, etc.)
+- Configurable key prefix for organizing objects
+- Automatic error logging
 
 ## Usage
 
@@ -42,21 +62,24 @@ from storage import Storage
 # Initialize S3 storage
 s3_storage = Storage("s3")
 
-# Save vCon data
+# Save vCon data (retrieves from Redis and stores in S3)
 s3_storage.save(vcon_id)
 
 # Retrieve vCon data
 vcon_data = s3_storage.get(vcon_id)
 ```
 
-## Implementation Details
+## Key Structure
+
+vCons are stored with keys following this pattern:
+```
+[s3_path/]YYYY/MM/DD/uuid.vcon
+```
 
-The S3 storage implementation:
-- Uses boto3 for AWS S3 operations
-- Implements retry logic
-- Supports multipart uploads
-- Provides encryption
-- Includes automatic metrics logging
+For example, a vCon created on January 15, 2024 with UUID `abc123` and `s3_path: vcons` would be stored at:
+```
+vcons/2024/01/15/abc123.vcon
+```
 
 ## Dependencies
 
@@ -65,15 +88,11 @@ The S3 storage implementation:
 
 ## Best Practices
 
-1. Secure credential management
-2. Implement proper access control
-3. Use appropriate storage classes
-4. Enable versioning
-5. Configure lifecycle rules
-6. Implement proper error handling
-7. Use appropriate encryption
-8. Monitor costs
-9. Implement retry logic
-10. Use appropriate regions
-11. Enable logging
-12. Regular backup verification 
+1. Always configure `aws_region` to match your bucket's region
+2. Use IAM roles with least-privilege access
+3. Enable bucket versioning for data protection
+4. Configure lifecycle rules for cost optimization
+5. Enable server-side encryption
+6. Use VPC endpoints for private connectivity
+7. Monitor with CloudWatch metrics
+8. Enable access logging for auditing

server/storage/s3/__init__.py

@@ -11,6 +11,31 @@
 default_options = {}
 
 
+def _create_s3_client(opts: dict):
+    """Create an S3 client with the provided options.
+
+    Required options:
+        aws_access_key_id: AWS access key ID
+        aws_secret_access_key: AWS secret access key
+
+    Optional options:
+        aws_region: AWS region (e.g., 'us-east-1', 'us-west-2')
+        endpoint_url: Custom endpoint URL for S3-compatible services
+    """
+    client_kwargs = {
+        "aws_access_key_id": opts["aws_access_key_id"],
+        "aws_secret_access_key": opts["aws_secret_access_key"],
+    }
+
+    if opts.get("aws_region"):
+        client_kwargs["region_name"] = opts["aws_region"]
+
+    if opts.get("endpoint_url"):
+        client_kwargs["endpoint_url"] = opts["endpoint_url"]
+
+    return boto3.client("s3", **client_kwargs)
+
+
 def save(
     vcon_uuid,
     opts=default_options,
@@ -19,11 +44,7 @@ def save(
     try:
         vcon_redis = VconRedis()
         vcon = vcon_redis.get_vcon(vcon_uuid)
-        s3 = boto3.client(
-            "s3",
-            aws_access_key_id=opts["aws_access_key_id"],
-            aws_secret_access_key=opts["aws_secret_access_key"],
-        )
+        s3 = _create_s3_client(opts)
 
         s3_path = opts.get("s3_path")
         created_at = datetime.fromisoformat(vcon.created_at)
@@ -45,11 +66,7 @@ def save(
 def get(vcon_uuid: str, opts=default_options) -> Optional[dict]:
     """Get a vCon from S3 by UUID."""
     try:
-        s3 = boto3.client(
-            "s3",
-            aws_access_key_id=opts["aws_access_key_id"],
-            aws_secret_access_key=opts["aws_secret_access_key"],
-        )
+        s3 = _create_s3_client(opts)
 
         s3_path = opts.get("s3_path", "")
         key = f"{s3_path}/{vcon_uuid}.vcon" if s3_path else f"{vcon_uuid}.vcon"