While operating a cluster, if every single transfer route fails for a particular virtual interface, a cluster can switchover to the other cluster. With this capability, the system can be recovered, without administrator's interference, by performing switchover within the cluster when detecting failures in the entire transfer route. Cluster fail-over is enabled in the initial setup for duplex transfer route operation in Fast switching mode, NIC switching mode, Virtual NIC mode, and GS linkage mode. This function is automatically configured when the cluster definition is defined.
Figure 2.44 Cluster failover due to line fault shows an example of fail-over to node B when communication is disabled via both eth0 and eth1 bundled with virtual interface sha0 on node A.
Information
The following is an example of Fast switching mode and this applies to NIC switching mode, Virtual NIC mode, and GS linkage mode as well.
In Fast switching mode, GLS determines that an error has occurred on a node when communication is cut off from another node on the same network (via dedicated monitoring frame) in Fast switching mode.
Note
When multiple virtual machines are created on one server to set up a cluster configuration, and Fast switching mode is used, an error will not occur to the cluster resources, even if a failure occurs to a switch which exists outside of the server. This is because a configuration is for successful monitoring at any time in the virtual switch in which multiple virtual machines are connected.
Information
In Fast switching mode, the system is set as a monitoring destination when it is added to the virtual interface bundles' networks.
In NIC switching mode, GLS determines that an error has occurred if HUB monitoring (by ping) fails a second time during standby patrol after the first HUB monitoring (by ping) failure and after NIC switching is performed.
Information
HUB monitoring is performed for the IP address that is set by the -p option of the hanetpoll command. For details, see "7.1 hanetconfig Command" and "7.7 hanetpoll Command".
In Virtual NIC mode, GLS determines that an error has occurred on a node if NIC is not recovered and monitoring error is detected also on a switched node after detection of a monitoring error on an active NIC and NIC switching.
Information
In network monitoring, HUB monitoring is performed for the IP address that is set by the -i option of the hanetpathmon command. In addition, setting up the standby patrol is not required. For details, see "7.1 hanetconfig Command" and "7.12 hanetpathmon Command".
In GS linkage mode, GLS determines that an error has occurred when every remote host monitoring (by ping) for the remote node and other nodes comprising the local cluster failed.
Information
Remote host monitoring is performed for the IP address that is set by the -t option of the hanetobserv command. For details, see "7.15 hanetobserv Command".