Unreliable deployment

This action releases a code version and immediately updates an alias to point to that version. This is never the desired outcome: new version takes some time to become available. This can result in failures similar to this one.

AWS CLI supports checks to facilitate preventing such premature deployments: lambda wait function-updated-v2 and lambda wait function-active-v2. This should be also doable via JS SDK.

Here's an AWS CLI version that prevents deploying unavailable functions:

set -euo pipefail
version=$(
    aws lambda update-function-code --publish \
    --function-name "$function" \
    --image-uri "$image" \
    | jq -r '.Version'
)
aws lambda wait function-updated-v2 --function-name "$function"
aws lambda wait function-active-v2 --function-name "$function"
aws lambda update-alias \
    --name "$alias_name" \
    --function-name "$function" \
    --function-version "$version"

Start at the action's deployment entry point and trace the update-function-code and alias-update sequence. Compare the AWS CLI waiters named in the issue with the JavaScript SDK flow, then verify that aliases are updated only after the function is updated and active. Done means deployments no longer point the alias at an unavailable version.