Details
-
Epic
-
Status: Complete
-
Medium
-
Resolution: Done
-
None
-
None
-
None
-
None
-
Switch off replicas
-
Unset
Description
We have a problem with backup instances thouse primaries are stopped. In this case primary nodes:
- replicas stashing messages and can't use it and dynamicly free memory
- replicas store all new requests and other replicas can't remove already oredered messages.
In RBFT we need in f+1 instances because if f primary (backup and master) nodes will malicious, one node will correct and pool will do view changes until the correct node becomes the primary.
If primary on backup malicious, the whole backup instance malicious and we don't need in this throughput metrics. And we can remove it.