Split Brain Syndrome Basic Concept in Oracle RAC

In Oracle RAC each node in the cluster is interconnected through a private interconnect. End-users connect to clusters through a public network. When a node is physically up and running and database instances are also running fine, but private interconnect fails between two or more nodes and an instance member fails to connect or ping to one another nodes, then the following scenario may occur.

è Instance think that the other instance that is unable to connect is down because of lack of connection between each other

è Meanwhile, both instances start working independently.

è Each RAC node can accept user connections and work independently.

è Ideally when Node1 is unable to ping to other nodes ex Node2 then Node2 will get evicted from the RAC groups to keep the data consistency and integrity.

Issue when both Instance start working independently

è Same data block might get read by both the instances server processes

è Might get written by both the instance server and background processes

è In both scenarios data integrity issue might occur as block changes by any instances will not get locked and recorded by other instances

The overall situation is termed as Split-Brain Syndrome in Oracle RAC.

Who will decide which node will survive?

è In a split-brain situation, the voting disk will determine which node(s) survive and which node(s) will be evicted.

è So voting disk concept plays a very important role in the split-brain situation.

Node Eviction Oracle Metalink Document- Note ID: 219361.1

Tim O'Reilly Gerald Venzl Gerald Venzl Jim Grisanzio Lucas Maldonado