Voting Disk
For a normal functioning of the cluster, every node should be able to communicate with every other node through cluster interconnect. Primary communication between nodes are through Cluster Interconnect. When there is a failure in reachingother nodes through interconnect, then voting disk is used to determine which nodes are part of cluster. Each node updates the voting file with information about the list of nodes it can reach through interconnect, and the list of nodes it is not able to reach through interconnect. This information is used to determine the list of nodes that can communicate among themselves. The remaining nodes that can not communicate are evicted out of the cluster. Eviction means that that node is rebooted or the clusterware stack on that node is rebooted.
Every node in cluster should access the shared storage only if it can communicate to other nodes via the cluster interconnect. Because cluster interconnect is used for co-ordinatedaccess to shared database files, to avoid any file corruption.
If the cluster interconnect communication fails, then each node can think that the other node is not available, and hence each node accesses the shared database files on its own without any co-ordination, thus corrupting the file. This is called split-brain situation in cluster environment. Voting Disk is used to avoid split-brain scenario in cluster environment.
For a normal functioning of the cluster, every node should be able to communicate with every other node through cluster interconnect. Primary communication between nodes are through Cluster Interconnect. When there is a failure in reachingother nodes through interconnect, then voting disk is used to determine which nodes are part of cluster. Each node updates the voting file with information about the list of nodes it can reach through interconnect, and the list of nodes it is not able to reach through interconnect. This information is used to determine the list of nodes that can communicate among themselves. The remaining nodes that can not communicate are evicted out of the cluster. Eviction means that that node is rebooted or the clusterware stack on that node is rebooted.
Every node in cluster should access the shared storage only if it can communicate to other nodes via the cluster interconnect. Because cluster interconnect is used for co-ordinatedaccess to shared database files, to avoid any file corruption.
If the cluster interconnect communication fails, then each node can think that the other node is not available, and hence each node accesses the shared database files on its own without any co-ordination, thus corrupting the file. This is called split-brain situation in cluster environment. Voting Disk is used to avoid split-brain scenario in cluster environment.
No comments:
Post a Comment