Improve Error Handling for Large Payloads

[samirrayani]

Describe the issue
When publishing a payload via SNS (and probably SQS too) that is too large, the response should provide a better warning/message. Instead a TypeError is thrown because there is no Error message in the response

Steps to reproduce
Steps to reproduce the behavior:

1. Create a new lambda and publish a large payload to it (arc.events.publish(...))
2. See error that says TypeError: @aws-lite/client: SNS.Publish: Cannot read properties of null (reading 'Error')

Expected behavior
A clear message indicating why the message was unable to be processed (in our case, too big!)

Screenshots

[brianleroux]

wow this is funny…

• max payload size for SNS is 256KB
• max payload size for SQS is 2GB

!!!

[samirrayani]

whoa maybe I should start using queues more? I would always just default to creating an event and add some concurrency but that's a pretty big diff

[brianleroux]

Well, it depends, queues are for for when you don't want concurrency. They are designed for throttling "first in first out". An example would be you need to call an API but are rate limited to 1 request per second. If you have two users making requests one will get throttled and fail…unless if you use a queue and enqueue requests to throttle them (on your side) to 1rps.

Events are the opposite. The pattern is "fan out" when you want to 1-many broadcast as fast as possible.

[samirrayani]

aha! that makes sense. thank you for the explanation 🤠