Amazon EventBridge

What you can build

• Replay a window of events after a fix

  An archive on the bus keeps events that match a pattern. After a target failure, the operator proposes a StartReplay for the affected time frame, limited to the rules that missed them. Replayed events carry a replay-name field.

  With Webhook Replay

• Route store and payment events to one operator

  Shopify and Stripe event sources are associated with a bus in your account. One rule sends their order and charge events to the connector's API destination, and the operator sees both sides of an order.

  With Order Risk

• Notice a route that has gone dark

  Events that exhaust the retry policy land in the rule's dead-letter queue. The operator reads that queue, names the target and the error code, and proposes a disable or a fix for your approval.

  With Integration Health, Vendor Outage

Requirements

• An AWS account with EventBridge in a Region where API destinations to public HTTPS endpoints are available
• IAM permission for PutEvents, PutRule, PutTargets, StartReplay, and the Describe and List actions the connector calls on the buses and rules in use
• A connection for the API destination; creating it adds the AmazonEventBridgeApiDestinationsServiceRolePolicy service-linked role, which manages the secret in Secrets Manager
• A public HTTPS endpoint with a publicly trusted certificate; API destinations do not support mutual TLS
• For SaaS events, an event source created on the vendor's side with your AWS account ID and then associated with an event bus

Authentication

AWS credentials signing requests with Signature Version 4 under an IAM principal permitted for the EventBridge actions used; each API destination's connection holds a Basic, OAuth client credentials, or API key secret in AWS Secrets Manager.

Limits

• A PutEvents request carries at most 10 entries, under 1 MB in total; JSON nested deeper than 1000 levels is not processed.
• A rule has at most 5 targets. A bus holds 300 rules by default, 100 in af-south-1 and eu-south-1, and only 30 may use wildcards.
• Requests to an API destination time out after 5 seconds; each destination is limited to 300 invocations per second per account per Region by default.
• Retries to an API destination stop after 24 hours or 185 attempts by default, and only 401, 407, 409, 429, and 5xx responses are retried.

Questions and answers

Does EventBridge retry a failed delivery to my endpoint?

Yes, for 401, 407, 409, 429, and 5xx responses, up to 24 hours and 185 attempts by default. It reads Retry-After and takes the more conservative wait. Other 4xx responses are not retried. Attach a dead-letter queue so exhausted events are kept rather than dropped.

Can I replay events that were already delivered?

Yes, if an archive on the source bus captured them. A replay selects a time frame and optionally specific rules, replays only to the source bus, and marks each event with a replay-name field. Ten replays can run at once per account per Region.

How do a SaaS vendor's events reach a bus?

The vendor creates an event source for your AWS account ID. You associate it with an event bus in the console, or with create-event-bus naming both name and event-source-name. Events published before the association exists are dropped and are not kept.