Is your feature request related to a problem? Please describe.
From the 2023 roadmap, delivery of data from our ITL applications have at-least once guarantees , but some users may need an exactly-once guarantee. Exactly-once guarantees are difficult to achieve but can be possible by utilizing caching in our sinks.
All sinks support a user-configured cache that enables idempotency
Describe alternatives you've considered
Idempotency is not usually required for metrics collection or log ingestion use cases since duplicate records are unlikely to cause adverse outcomes -- in fact, the vast majority of systems like Substation use at-least once guarantees because they are affordable (from a CPU / cost perspective) and low impact when errors occur. Users should make themselves aware of the risks posed by systems that use at-least once guarantees and plan accordingly. Useful information on guarantees can be found in this blogpost.
Is your feature request related to a problem? Please describe.
From the 2023 roadmap, delivery of data from our ITL applications have at-least once guarantees , but some users may need an exactly-once guarantee. Exactly-once guarantees are difficult to achieve but can be possible by utilizing caching in our sinks.
It's worth mentioning that exactly-once guarantees require both consumers and producers to support idempotency, but this issue is focused on producers. None of the serverless cloud services that Substation uses today supports idempotency; for example, the Kinesis service describes idempotency as a task that must be handled by consumers and producers.
Describe the solution you'd like
Add support for the following:
Describe alternatives you've considered
Idempotency is not usually required for metrics collection or log ingestion use cases since duplicate records are unlikely to cause adverse outcomes -- in fact, the vast majority of systems like Substation use at-least once guarantees because they are affordable (from a CPU / cost perspective) and low impact when errors occur. Users should make themselves aware of the risks posed by systems that use at-least once guarantees and plan accordingly. Useful information on guarantees can be found in this blogpost.
Additional context N/A