98 IBM System Storage N series Hardware Guide7.4.4 HA pair configuration failover basic operationsWhen a failover occurs, the running partner node in the HA pair configuration takes over thefunctions and disk drives of the failed node. It does so by creating an emulated storagesystem that runs the following tasks: Assumes the identity of the failed node. Accesses the disks of the failed node and serves their data to clients. The partner node maintains its own identity and its own primary functions, but alsohandles the added function of the failed node through the emulated node.7.4.5 Connectivity during failoverBoth front-end and back-end operations are affected during a failover. On the front end arethe IP addresses and the host name. On the back end, there is the connectivity andaddressing to the disk subsystem. Both the back-end and front-end interfaces must beconfigured correctly for a successful failover.Reasons for HA pair configuration failoverThe conditions under which takeovers occur depend on how you configure the HA pairconfiguration. Takeovers can be initiated when one of the following conditions occurs: An HA pair node that is configured for immediate takeover on panic undergoes a softwareor system failure that leads to a panic. A node that is in an HA pair configuration undergoes a system failure (for example,NVRAM failure) and cannot reboot. There is a mismatch between the disks that one node can see and the disks that the othernode can see. One or more network interfaces that are configured to support failover becomesunavailable. A node cannot send heartbeat messages to its partner. This situation might happen if thenode experienced a hardware failure or software failure that did not result in a panic, butstill prevents it from functioning correctly. An example is a failure in the interconnect cable. You halt one of the HA pair nodes without using the -f flag. The -f flag applies only tostorage systems in an HA pair configuration. If you enter the halt -f command on anN series, its partner does not take over. You initiate a takeover manually.Failover because of disk mismatchCommunication between HA pair nodes is first established through the HA pair configurationinterconnect adapters. At this time, the nodes exchange a list of disk shelves that are visibleon the A loop and B loop of each node. If the B loop shelf count on its partner is greater thanRemember: When a failover occurs, existing CIFS sessions are terminated. A gracefulshutdown of the CIFS sessions is not possible, and some data transfers might beinterrupted.Restriction: If the storage for a node also loses power at the same time, a standardtakeover is not possible.
