Document your server configuration using the following guidelines:
- Determine the server names, processor types, memory and other I/O devices for your configuration. When you specify a backup server, you should ensure that the server you select has the capacity to perform the processing should a failure occur on the primary server.
- Determine your communications connection requirements.
Important: Potentially, clustered configurations have two types of communications requirements: cluster requirements and user requirements.
- Cluster – A LifeKeeper cluster requires at least two communication paths (also called “comm paths” or “heartbeats”) between servers. This redundancy helps avoid “split-brain” scenarios due to communication failures. Two separate LAN-based (TCP) comm paths using dual independent subnets are recommended, and at least one of these should be configured as a private network. Using a combination of TCP and TTY is also supported. A TTY comm path uses an RS-232 null-modem connection between the servers’ serial ports.
Note that using only one comm path can potentially compromise the ability of systems in a LifeKeeper cluster to communicate with each other. If a single comm path is used and the comm path fails, then LifeKeeper hierarchies may try to come into service on multiple systems simultaneously. This is known as a false failover or a “split-brain” scenario. In the “split-brain” scenario, each server believes it is in control of the application and thus may try to access and write data to the shared storage device. To resolve the split-brain scenario, LifeKeeper may cause servers to be powered off or rebooted or leave hierarchies out-of-service to assure data integrity on all shared data. Additionally, heavy network traffic on a TCP comm path can result in unexpected behavior, including false failovers and the failure of LifeKeeper to initialize properly.
- User – We recommend that you provide alternate LAN connections for user traffic – that is, a separate LAN connection than the one used for the cluster heartbeat. However, if two TCP comm paths are configured (as recommended), one of those comm paths can share the network address with other incoming and outgoing traffic to the server.
- Identify and understand your shared resource access requirements. Clusters that use shared storage can utilize either shared SCSI buses or Fibre Channel loops. Because LifeKeeper locks resources to one server, you must ensure that only one server requires access to all locked resources at any given time. LifeKeeper device locking is done at the Logical Unit (LUN) level. For active/active configurations, each hierarchy must access its own unique LUN. All hierarchies accessing a common LUN must be active (in-service) on the same server.
- Determine your shared memory requirements. Remember to take into account the shared memory requirements of third-party applications as well as those of LifeKeeper when configuring shared memory and semaphore parameters. See Tuning in Technical Notes for LifeKeeper’s shared memory requirements.
Sample Configuration Map for LifeKeeper Pair
This sample configuration map depicts a pair of LifeKeeper servers sharing a disk array subsystem where, normally, Server 1 runs the application(s) and Server 2 is the backup or secondary server. In this case, there is no contention for disk resources because one server at a time reserves the entire disk storage space of the disk array. The disk array controller is labeled “DAC,” and the SCSI host adapters (parallel SCSI, Fibre Channel, etc.) are labeled “SCSI HA.”
A pair of servers is the simplest LifeKeeper configuration. When you plan a cluster consisting of more than two servers, your map is even more critical to ensure that you have the appropriate connections between and among servers. For example, in a multi-directional failover configuration, it is possible to define communications paths within LifeKeeper when the physical connections do not exist. Each server must have a physical communication path to every other server in the cluster in order to provide cascading failover capability.