Fail-over and Split Brain Policies
Fail-Over Policy
The SMI Cluster Manager implements the following policies during fail-over and split brain scenarios.
During a fail-over, the active node shuts down all the K8s and Docker services. The DRDB disk is unmounted and demoted when all the services are stopped.
The standby node promotes and mounts the DRDB disk and starts the Docker and K8s services.
Split Brain Policy
The following policies are defined for automatic split brain recovery:
-
discard-least-changes - This policy is implemented when there is no primary node. It discards and rolls back all the modifications on the host where fewer changes have occurred.
-
discard-secondary - This policy is implemented when there is a primary node. It makes the secondary node the split brain victim.