Failover clusters or High availability clusters, or Server Clusters are clusters of computer nodes configured to take over each others functionalities when the primary node fails. A heartbeat private network connection is used to monitor the health and status of each node in the cluster and cluster software detects hardware/software faults and restarts the application in another node in the cluster, with least amount of downtime for the application. This makes for high availability and is recommended for organizations that run mission critical or client facing applications such file sharing applications on a network, business applications and customer service applications that have to be accessibleFailover clusters or High availability clusters, or Server Clusters are clusters of computer nodes configured to take over each others functionalities when the primary node fails.
A heartbeat private network connection is used to monitor the health and status of each node in the cluster and cluster software detects hardware/software faults and restarts the application in another node in the cluster, with least amount of downtime for the application. This makes for high availability and is recommended for organizations that run mission critical or client facing applications such file sharing applications on a network, business applications and customer service applications that have to be accessibleFailover clusters or high availability clusters or server clusters are clusters of computer nodes that are configured to take over each others functionality when one of them fails.
The cluster software detects a hardware or software failure and restarts the application in another node of the cluster with least amount of downtime. This makes for high availability of computing time and is recommended for organizations which run mission critical or client facing applications that must be operational 24 x 7 x 365.
A minimum of two computers make up a cluster. However, large organizations have multiple nodes in a cluster to provide adequate redundancy. The configurations of these clusters are categorized as Active/Active, Active/Passive, N+1, N+M, N-to-1 or N to N clusters. When the nodes use homogeneous software combination and pass on the traffic uniformly balanced across the cluster from a failed node, it is known as an Active/Active cluster. The Active/Passive cluster has one active and one passive node and the latter becomes active when the former fails. In N+1 cluster heterogeneous services run simultaneously on different nodes of the cluster and when one node fails an extra node is brought in to assume the role of any one of the primary nodes which has failed. N+M clusters refer to clusters which have multiple services running on different nodes of the clusters, but have only one dedicated failover node of insufficient redundancy and several standby servers included in the cluster. In N-to-1 clusters the standby node is allowed to assume the active role until the original is restored. However, the services and instances are required to be ‘failed-back’ to the original node to make it highly available. The N-to-N is a combination of the Active/Active and N+M clusters.
Failover clusters or High availability clusters, or Server Clusters are clusters of computer nodes configured to take over each others functionalities when the primary node fails. A heartbeat private network connection is used to monitor the health and status of each node in the cluster and cluster software detects hardware/software faults and restarts the application in another node in the cluster, with least amount of downtime for the application. This makes for high availability and is recommended for organizations that run mission critical or client facing applications such file sharing applications on a network, business applications and customer service applications that have to be accessible.
The services provided by the cluster are accessed via a network address known as the logical host or cluster logical host. This address is not identified with any single node in the cluster but linked with the service provided by the cluster.
Applications that run in a high availability environment must satisfy certain basic conditions. The application must be capable of being easily started, stopped, force-stopped and checked for status. It must have a command line interface and scripts to control it and it must be capable of supporting multiple instances. The application must use shared storage such as NAS or SAN. The application must store its state in largely non-volatile storage and have the capability of restarting on another node in the state at which the original node went down. More importantly the application on restart must not corrupt the data of the saved state.
As stated above high availability clusters are monitored for the status of their nodes by a heartbeat private network connection. When one node breaks down, another node takes over the application. However, in certain circumstances the clustering software may engender a situation that is known as split brain. Split brain happens when all the private network links go down simultaneously and the software detects that the cluster nodes are still running. The nodes in the cluster assume that the other nodes are down and will start an instance of the services the other nodes are running. This will result in multiple copies of the application running simultaneously. This could lead to corruption of data.
Failover clustering is intended to provide hardware redundancy. It does not improve performance; protect systems against viruses, denial of services, database corruption, and logical corruption or software failures. It is not a load balancing solution and should not be used as such.
by Dr.Vanitha Vaidialingam alias Samantha
