"health"property of the common JSON object or as an argument in the StackState CLI when health data is sent to StackState. The supported models are:
REPEAT_SNAPSHOTSconsistency model works with periodic, full snapshots of all checks in an external monitoring system. StackState keeps track of the checks in each received snapshot and decides if associated external check states need to be created, updated or deleted in StackState. For example, if a check state is no longer present in a snapshot. This model offers full control over which external checks will be deleted as all decisions are inferred from the received snapshots. There is no ambiguity over the external checks that will be present in StackState.
REPEAT_STATESconsistency model works with periodic checks received from an external monitoring system. StackState keeps track of the checks and decides if associated external checks need to be created or updated in StackState. A configurable expiry mechanism is used to delete external checks that are not observed anymore. This model offers less control over data than the
REPEAT_SNAPSHOTSmodel. As an expiry configuration is used to delete external checks, it might happen that elements are deleted due to barely missing the expiry timeout. This would reflect as external checks disappearing and reappearing in StackState.
TRANSACTIONAL_INCREMENTSconsistency model is designed to be used on streaming systems where only incremental changes are communicated to StackState. As there is no repetition of data, data consistency is upheld by ensuring that at-least-once delivery is guaranteed across the entire pipeline. To detect whether any data is missing, StackState requires that both a checkpoint and the previous checkpoint are communicated together with the
check_states. This model requires strict control across the whole pipeline to guarantee no data loss.
sub_stream_idcan be omitted from the health payload. StackState will assume that all the external health checks belong to a single, default sub stream.