A Lambda Errors alarm watches a function that runs only a few times each day. The team wants missing datapoints to be treated as healthy, while real nonzero error datapoints should still trigger the configured threshold. What should the engineer change?

1. Set the alarm missing-data treatment to ignore to guarantee an OK state during inactivity.
2. Set the alarm missing-data treatment to breaching.
3. Set the alarm missing-data treatment to notBreaching.
4. Disable the alarm actions so that missing datapoints are converted into zero-valued metric samples.