Cluster IP failover

If the configuration node fails, the cluster IP addresses are transferred to a new node. The cluster services are used to manage the transfer of the cluster IP addresses from the failed configuration node to the new configuration node.

The following changes are performed by the cluster service:

If software on the failed configuration node is still operational, the software shuts down the cluster IP interfaces. If the software cannot shut down the cluster IP interfaces, the hardware service forces the node to shut down.
When the cluster IP interfaces shut down, all remaining nodes choose a new node to host the configuration interfaces.
The new configuration node initializes the configuration daemons, including sshd and httpd, and then binds the cluster IP interfaces to its Ethernet ports.
The router is configured as the default gateway for the new configuration node.
The routing tables are established on the new configuration node for the cluster IP addresses. The new configuration node sends five unsolicited address resolution protocol (ARP) packets for each IP address to the local subnet broadcast address. The ARP packets contain the cluster IP and the media access control (MAC) address for the new configuration node. All systems that receive ARP packets are forced to update their ARP tables. Once the ARP tables are updated, these systems can connect to the new configuration node.
Note: Some Ethernet devices might not forward ARP packets. If the ARP packets are not forwarded, connectivity to the new configuration node cannot be established automatically. To avoid this problem, configure all Ethernet devices to pass unsolicited ARP packets. You can restore lost connectivity by logging into the SAN Volume Controller and starting a secure copy to the affected system. Starting a secure copy forces an update to the ARP cache for all systems connected to the same switch as the affected system.

Ethernet link failures

If the Ethernet link to the SAN Volume Controller cluster fails because of an event unrelated to the SAN Volume Controller, such as a cable being disconnected or an Ethernet router failure, the SAN Volume Controller does not attempt to failover the configuration node to restore cluster IP access. SAN Volume Controller provides the option for two Ethernet ports, each with its own management IP address, to protect against this type of failure. If you cannot connect through one IP address, attempt to access the cluster through the alternate IP address.

Note: IP addresses that are used by hosts to access the cluster over an Ethernet connection are different from cluster IP addresses.

Routing considerations for event notification

SAN Volume Controller supports the following protocols that make outbound connections from the cluster:

E-mail
Simple Network Mail Protocol (SNMP)
Syslog
Network Time Protocol (NTP)

One or more of these protocols can be configured on the cluster to receive event notifications. When making outbound connections, the SAN Volume Controller uses the following routing decisions:

If the destination IP address is in the same subnet as one of the cluster IP addresses, then SAN Volume Controller sends the packet immediately.
If the destination IP address is not in the same subnet as either of the cluster IP addresses, then SAN Volume Controller sends the packet to the default gateway for Ethernet port 1.
If the destination IP address is not in the same subnet as either of the cluster IP addresses and Ethernet port 1 is not connected to the Ethernet network, then SAN Volume Controller sends the packet to the default gateway for Ethernet port 2.

When configuring any of these protocols for event notifications, use these routing decisions to ensure error notification works correctly in the event of a network failure.