I don’t understand the difference between the 2. Can someone explain? I’ve read about the rules which govern if the partition is active or not? I don’t understand what happens if the partition becomes unavailable in the sub-clusters(specially when it becomes unavailable in all sub-clusters). Is it possible to revive it like in case of dead partitions?
The unavailable_partitions metric means there are inaccessible partitions and nodes from the roster are missing from the cluster. So they could become available if the missing nodes return.
The dead_partitions metric are unavailable_partitions when all roster nodes are present. It implies that RF or more nodes had an unsafe shutdown or storage wiped i.e. potential data loss. As an operator, you may have a recovery procedure (such as stopping all traffic, starting reviving the cluster and restoring from a backup). To acknowledge the potential data loss the operator will need to issue the revive command.
1 Like