The DevOps workshop at the canary deployment step seems to have a misconfiguration for the underlying metric. The workshop guides you to set up an erroring endpoint and then set up a canary alarm to roll back deployments when errors occur. Then it has you set up a local script to ping the endpoint repeatedly during deployment to trigger the rollback. However, after following the steps, my deployment succeeded unexpectedly.
This seems to be due to a misconfiguration on the underlying canary alarm metric. It shows "Insufficient data" even though the application is returning errors.
The DevOps workshop at the canary deployment step seems to have a misconfiguration for the underlying metric. The workshop guides you to set up an erroring endpoint and then set up a canary alarm to roll back deployments when errors occur. Then it has you set up a local script to ping the endpoint repeatedly during deployment to trigger the rollback. However, after following the steps, my deployment succeeded unexpectedly.
This seems to be due to a misconfiguration on the underlying canary alarm metric. It shows "Insufficient data" even though the application is returning errors.
The metric alarm has the following details:
I did some ad-hoc testing and recreated the metric with the region removed from the dimensions:
and it seemed to fix the issue. This metric ended up picking up all of the errors.
Is this an underlying issue with the canary configuration in SAM? Or am I missing something here?