Cluster Heartbeats

In a replicated data cluster, robust heartbeating is accomplished through dual, dedicated networks over which the Low Latency Transport (LLT) runs. Additionally, you can configure a low-priority heartbeat across public networks.

In a global cluster, network heartbeating is accomplished by sending ICMP pings over the public network between the two sites. VCS global clusters minimize the risk of split-brain by sending ICMP pings to highly available IP addresses and by notifying administrators when the sites cannot communicate.

In global clusters, the VCS Heartbeat agent sends heartbeats directly between the Symmetrix arrays, given the Symmetrix ID of each array. This heartbeat offers the following advantages:

• VCS does not mistakenly interpret the loss of ICMP heartbeats over the public network as a site failure because the Symmetrix heartbeat shows that the arrays are alive.
• If the loss of heartbeats occurs due to the failure of all hosts in the primary cluster, a failover may be required even if the array is alive. In any case, it is important to distinguish between a host-only crash and a complete site failure. In a host-only crash, only the ICMP heartbeat signals a failure via an SNMP trap. No cluster failure notification occurs because a surviving heartbeat exists. This trap is the only notification to fail over an application.
• The heartbeat is then managed completely by VCS and reports as being down only when the remote array is not visible by the symrdf ping command.