Cloverleaf Cluster
Clustering provides high availability by enabling application failover. The state of each individual node is periodically monitored and automatically relocates the application from a failed node to a designated secondary node.
When a failover happens, there might be a brief delay. In this case, refresh after the failover is finished.
Clustering can significantly reduce downtime and increase productivity by providing highly available service to all users.
Cluster has these key components:
- Hardware:
- SAN (shared storage between cluster nodes)
- Private communication for heartbeats and cluster interconnect
- Public network for user access
- Local disks (unshared storage)
- Resource Group Manager
- CloverHA Agent
- Software:
- OS (AIX,SUN OS, HP-UX, RHEL, and Windows)
- Clustering Software (PowerHA, Sun Cluster, MC-Service Guard, Redhat Cluster, Windows Clustering and Veritas)
- SAN Replication (IBM Metro Mirror, EMC Recover point and HP Storage works Multi-site Replication solution)
- SAN: Storage Area Network
This consists of disks that can be connected to more than one node at a time. Disks are made highly available through SAN. It can tolerate single node failure, since a physical path to the disks exists through the designated secondary node. It provides disk mirroring and stripping which protects against individual disk and node failures.
- Resource Group Manager
This provides mechanisms for high availability. This is policy-driven and automatically stops and starts the application on selected nodes. In the event of a node failure, it stops the application on a failed node and starts the application on the designated secondary node.
- CloverHA Agent
This is a dynamic agent with no user intervention. It is implemented as a shell script that gives maximum flexibility to users.
The Cloverleaf Cluster types are:
- Active-Passive
- Active-Test
- Active-Active
- Geo-Cluster
Active-Passive
In this configuration, node1 hosts Cloverleaf and node2 is designated as a standby node, waiting for node1 to fail.
When a failover happens, Cluster takes these actions:
- Moves the application data files to the secondary node.
- Moves the application IP address to the secondary node.
- Starts the application on the secondary node. If Clients notice a brief interruption in service, then they must refresh/reconnect.
Active-Test
In this configuration, node1 is hosting a production instance of Cloverleaf. Node2 is designated as the standby node and simultaneously is hosting the test instance.
When a failover happens, Cluster takes these actions:
- Moves the production data files to the secondary/test node.
- Moves the production IP address to the secondary/test node.
- Starts the production application on the secondary node. If Clients notice a brief interruption in service, then they must refresh/reconnect.
Active-Active
In this configuration, node1 is hosting the production instance of Cloverleaf. Node2 is hosting the second instance of Cloverleaf. Both nodes act as secondary to each other, also known as "mutual-take-over". When a failover happens, node2 becomes double-active.
Cluster takes these actions:
- Moves the production data files to the secondary node.
- Moves the production IP address to the secondary node.
- Clustering Software (PowerHA, Sun Cluster, MC-Service Guard, Redhat Cluster, WindowsStarts production application on the secondary node. If Clients notice a brief interruption in service, then they must refresh/reconnect.
Geo-Cluster
In this configuration, the cluster is configured across two data centers. This provides a greater level of availability and protection of storage by having a second copy of the data at the secondary location. When the primary data center fails, the application is relocated to the secondary/remote data center.
Cluster takes these action:
- Designates the remote data center as a primary location.
- Moves the application IP to the remote data center.
- Starts the application at the remote data center. If Clients notice a brief interruption in service, then they must refresh/reconnect.