Currently errors encountered during the process of handling a message (from PubSub, etc) are sent directly back to the user via the response topic.
Consider:
Resending failed requests back to backends inside of the message handler function.
Nack'ing messages and relying on messaging features to handle retries (PubSub Expo-Backoff)
Sending requests back to the messaging system allows for other clusters to attempt the retry. It might make sense to consider how to retry (local-in-handler? re-queue? both? neither?) on a error-case-by-case basis.
Currently errors encountered during the process of handling a message (from PubSub, etc) are sent directly back to the user via the response topic.
Consider:
Sending requests back to the messaging system allows for other clusters to attempt the retry. It might make sense to consider how to retry (local-in-handler? re-queue? both? neither?) on a error-case-by-case basis.