[jennyhliu]

Summary: Summary of changes

Addresses CUMULUS-5731: Provision a dedicated custom event bus per stack

Changes

• Provisioned a dedicated custom event bus per stack and updated the Cumulus workflow event rule and targets to use it.

PR Checklist

• Update CHANGELOG
• Unit tests
• Ad-hoc testing - Deploy changes and test manually
• Integration tests

---

📝 Note:
For most pull requests, please Squash and merge to maintain a clean and readable commit history.

[npauzenga]

This looks great @jennyhliu. A couple questions but I see no reason this won't work.

• The scope here includes the workflow events but not scheduled/cron rules (thanks for the detailed ADR). Do we have work scoped to switch the remaining rules?

• Have you been able to verify in a deployed stack that sf events are directed to the appropriate bus? We can't really automate tests here so just verifying.

• As far as cost, do we have a sense of how many events we'll be forwarding? I see your note that it's $1.00/million events. The limit we're avoiding is the 2,000 rules per event bus but I'm not sure how this translates to cost. Likely not much.

[npauzenga]

Approving because I don't see blocking issues.

[jennyhliu]

This looks great @jennyhliu. A couple questions but I see no reason this won't work.

• The scope here includes the workflow events but not scheduled/cron rules (thanks for the detailed ADR). Do we have work scoped to switch the remaining rules?
  Yes, we have CUMULUS-5732 to migrate legacy scheduled rules to EventBridge Scheduler.
• Have you been able to verify in a deployed stack that sf events are directed to the appropriate bus? We can't really automate tests here so just verifying.
  I verified that only one rule is created on default event bus for my stack, and remaing rules are created on the dedicated custom event bus for this stack.
  In aws console, rule jl-rds-tf-sfn-execution-forwarder monitoring tab shows "Matched Events count", event bus jl-rds-tf-cumulus-event-bus monitoring tab shows "Successful invocation attempts count".
  Also sfEventSqsToDbRecords lambda consumes records written by the dedicated custom event bus, thus integration tests pass.
• As far as cost, do we have a sense of how many events we'll be forwarding? I see your note that it's $1.00/million events. The limit we're avoiding is the 2,000 rules per event bus but I'm not sure how this translates to cost. Likely not much.
  One workflow execution will have 2 state change, and 2 events if less than 64KB size, each 64 KB chunk of a payload is billed as one event. If a granule ingest will have 10 events, 1 million granules will cost $10 for forwarding events.

[npauzenga]

Thanks for the reply @jennyhliu. Let's go ahead and merge