AWS S3 Event

The AWS S3 Event Receiver consumes S3 event notifications for object creation events (s3:ObjectCreated:*) and emits the S3 object as the string body of a log record.

Supported Platforms

Platform	Supported
Linux	✓
Windows	✓
macOS	✓

Available in the Bindplane Distro for OpenTelemetry Collector v1.74.0+.

Prerequisites

• An AWS account with access to S3 and SQS.
• An SQS queue configured to receive S3 event notifications.

1. Configure an S3 bucket to send event notifications to an SQS queue for object creation events.
   • Configure your S3 event notifications with BatchSize: 1 to ensure each SQS message contains only one S3 event.
   • This setting is crucial because if an object cannot be accessed (e.g., 404 error), the entire SQS message is preserved for retry.
   • If a message contains multiple objects and one fails, all objects will be reprocessed on retry, causing unnecessary duplication.
2. Ensure the collector has permission to read and delete messages from the SQS queue.
3. Ensure the collector has permission to read objects from the S3 bucket.

How It Works

1. The receiver polls an SQS queue for S3 event notifications.
2. When an object creation event (s3:ObjectCreated:*) is received, the receiver downloads the S3 object.
3. The receiver reads the object into the body of a new log record.
4. Non-object creation events are ignored but removed from the queue.
5. If an S3 object is not found (404 error), the corresponding SQS message is preserved for retry later.

Configuration Fields

Field	Type	Default	Required	Description
sqs_queue_url	string		true	The URL of the SQS queue to poll for S3 event notifications.
standard_poll_interval	duration	15	false	The interval (in seconds) at which the SQS queue is polled for messages.
max_poll_interval	duration	2	false	The maximum interval (in seconds) at which the SQS queue is polled for messages.
polling_backoff_factor	float	2	false	The factor by which the polling interval is multiplied after an unsuccessful poll.
workers	int	5	false	The number of workers to process events.
visibility_timeout	duration	300	false	How long (in seconds) messages received from the queue will be invisible to other consumers.
max_log_size	int	1048576	false	The maximum size of a log record in bytes. Logs exceeding this size will be split
max_logs_emitted	int	1000	false	The maximum number of log records to emit in a single batch. A higher number will result in fewer batches, but more memory.

Example Configuration