You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Stream & Pipeline consumers should be able to nack an event with a delay.
Stream controllers will need to be updated to use a different tracking mechanism for group offsets.
Delayed items will need to be recorded as outstanding and recorded on disk (just the offset).
If a controller restarts, it will need to be able to detect such delayed events for a group and then spawn them as timer delays until they can be delivered.
Pipeline controllers will need to be updated to not block max parallel consumers when there are no consumers for a particular stage, as this would cause all pipeline processing to halt for a partition just because a stage consumer is not active.
The metrics system should still alert on such cases.
The way we model such a delay for a Pipeline stage will be different, but we should be able to just record this info under a stage specific key for the pipeline.
The text was updated successfully, but these errors were encountered:
The text was updated successfully, but these errors were encountered: