Improving cluster network heartbeating resiliency for Db2 HADR with Pacemaker

Summary

Corosync Cluster Engine’s main role in the Db2 Automated Failure HADR solution with Pacemaker is to provide the cluster membership information to the cluster manager to carry out the appropriate actions and ensure all managed resources are in their respective wanted states. Corosync employs the concept of heartbeat rings where TCP/IP capable adapters in each cluster host participate in exchanging messages among them. One network interface from each host forms a ring for heart-beating purposes. A host is deemed inaccessible only if all local network interfaces fail to exchange messages in all rings it participates in. The more rings a host participates in, the fewer false positives it generates, and the reliability of this node-liveliness test improves.

Before Db2 V11.5.6.0, the network interface associated with the hostname is the default interface participating in the node liveliness test. This configuration is done automatically during the cluster setup. While instructions are available in the open source Corosync documentation to set up more heartbeat rings, it is not tailored toward the Db2 Automated Failover HADR solution with Pacemaker specifically. This technical document intends to provide validated step-by-step instructions to set up multiple heartbeat rings for a Pacemaker HADR cluster after the initial cluster setup.