Architecture considerations and best practices for architecting an Oracle RAC solution with Serviceguard and SGeRAC

15
For faster cluster recovery:
• Use a common, dedicated HA network for Serviceguard heartbeat, CFS interconnect (if CFS is
used), CSS heartbeat, and RAC interconnect.
• Use Serviceguard Local LAN failover for network HA.
• Configure the Serviceguard heartbeat on the same network as the CSS heartbeat. CSS only
supports one active network and Serviceguard heartbeat will fail and cause the node to be evicted
if the network used by CSS fails. If you do not do this, it will take a CSS timeout of 600 seconds
before node failure and recovery will occur.
Configuration Option 1a
For some configurations, it is preferable to set up redundant, active Serviceguard heartbeat networks.
The second Serviceguard heartbeat provides additional robustness and enables faster cluster
reformation time, leading to improved availability.
When sharing a common network for Serviceguard heartbeat, CFS interconnect (if CFS is used),
Oracle Clusterware CSS heartbeat, and RAC interconnect, heavy RAC interconnect traffic may
interfere with cluster heartbeat traffic, potentially causing heartbeat message latency to exceed the
configured MEMBER_TIMEOUT. This can lead to unnecessary cluster reformations, and even nodes
being evicted from the cluster, in high-traffic scenarios.
As compared to Option 1, this configuration provides protection against high Cache Fusion traffic
interfering with the Serviceguard heartbeat, by way of the second, redundant Serviceguard heartbeat.
A recommended practice is to configure cluster interconnect monitoring to provide additional
robustness for monitoring the CSS heartbeat.