This guide covers what changes when upgrading from v1.0 to v1.1. There are no breaking changes and no new migrations, but there are behavior changes and deprecations to review before upgrading.

What's New

Configurable webhook system header names

You can now pin the exact name of each system header the webhook destination sends, instead of only controlling the shared prefix:

• DESTINATIONS_WEBHOOK_EVENT_ID_HEADER_NAME
• DESTINATIONS_WEBHOOK_TIMESTAMP_HEADER_NAME
• DESTINATIONS_WEBHOOK_TOPIC_HEADER_NAME
• DESTINATIONS_WEBHOOK_SIGNATURE_HEADER_NAME

Unset keeps the default <prefix><key> behavior, an explicit value pins that exact name, and an empty string disables the header. See Webhook destination for details, including the duplicate-name validation rules.

IAM role authentication for AWS SQS

AWS_SQS_ACCESS_KEY_ID and AWS_SQS_SECRET_ACCESS_KEY are now optional. When omitted, Outpost uses the AWS SDK default credential chain, so you can authenticate with an IAM role (for example EKS Pod Identity, IRSA, or an instance profile) instead of static credentials. See Configuration.

Delivery attempt operator events

Two new operator events are published: attempt.success and attempt.failed, emitted once per delivery attempt. Existing operator event topics are unchanged.

Faster log processing

Alert evaluation and operator event delivery during log processing now run in parallel per destination instead of serially. This is an internal throughput improvement with no configuration changes.

Behavior Changes

Retry tasks now dead-letter instead of retrying forever

Before v1.1, a delivery retry task that repeatedly failed to process (for example, a malformed task payload) was redelivered indefinitely and could occupy a delivery worker forever.

Starting in v1.1, the retry scheduler:

1. Applies exponential backoff between receives of a failing task (doubling per receive, capped at 15 minutes).
2. After 5 receives, moves the task to a dead-letter queue named deliverymq-retry-dlq, stored in Redis alongside the retry queue itself.

Each dead-lettered task logs an error, task exceeded max receive count, moved to dead-letter queue, that includes the queue name, message ID, receive count, and the full task payload. Alert on this log line to catch dead-lettered tasks.

There is no automatic redrive. Dead-lettered tasks stay in Redis under the rsmq:deliverymq-retry-dlq keys (prefixed with your DEPLOYMENT_ID if set) for manual inspection.

Stored response body size cap

Outpost stores each destination response body on the delivery attempt. Starting in v1.1, bodies larger than DESTINATIONS_WEBHOOK_MAX_RESPONSE_BODY_BYTES (default 131072, 128 KiB) are replaced with a placeholder instead of stored. This prevents oversized attempt logs from exceeding the message queue's per-message size limit, which previously caused those deliveries to be retried indefinitely.

To restore the previous unbounded behavior, set DESTINATIONS_WEBHOOK_MAX_RESPONSE_BODY_BYTES=0. See Webhook destination for sizing guidance.

Deprecations

The DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_*_HEADER flags are deprecated and will be removed in a future version. They still work in v1.1, but a set *_HEADER_NAME variable takes precedence over the matching flag.

Deprecated	Replacement
DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_EVENT_ID_HEADER=true	DESTINATIONS_WEBHOOK_EVENT_ID_HEADER_NAME=""
DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_TIMESTAMP_HEADER=true	DESTINATIONS_WEBHOOK_TIMESTAMP_HEADER_NAME=""
DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_TOPIC_HEADER=true	DESTINATIONS_WEBHOOK_TOPIC_HEADER_NAME=""
DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_SIGNATURE_HEADER=true	DESTINATIONS_WEBHOOK_SIGNATURE_HEADER_NAME=""

The deprecated fields are also removed from the OpenAPI spec and generated SDKs; the environment variables and Config API keys continue to work until removal.

Upgrade Checklist

1. Before upgrading:

   • [ ] If you use any DESTINATIONS_WEBHOOK_DISABLE_DEFAULT_*_HEADER flag, switch to the *_HEADER_NAME="" replacement
   • [ ] If your destinations return response bodies larger than 128 KiB and you rely on them being stored, set DESTINATIONS_WEBHOOK_MAX_RESPONSE_BODY_BYTES accordingly (or 0 for no cap)
   • [ ] If OPERATION_EVENTS_TOPICS=*, decide whether your sink can handle per-attempt volume; otherwise replace * with an explicit topic list

2. Upgrade:

   • [ ] Update Outpost to v1.1 (no new migrations in this release)

3. After upgrading:

   • [ ] Add an alert on the task exceeded max receive count, moved to dead-letter queue error log
   • [ ] Check deliverymq-retry-dlq for tasks that were previously stuck in a retry loop

Thanks

This release cycle includes first-time contributions from @LendritIbrahimi, @ambroziepaval, @mvanhorn, @lovlyx, @distroaryan, and @Abdulmumin1. Thank you for your contributions!