Details
-
Sub-task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
-
None
Description
We would like to move to a push model for sending committed data to the state machine. This simplifies the state machine a bit since it does not need to track its own position and poll for new data. It also allows the raft layer to ensure that all committed data up to the state of a leader epoch has been sent before allowing the state machine to begin sending writes. Finally, it allows us to take advantage of optimizations. For example, we can save the need to re-read writes that have been sent to the leader; instead, we can retain the data in memory and push it to the state machine after it becomes committed.
Attachments
Issue Links
- links to