12.5 NIFF Revisited

NIFF is automatically configured in a cluster because the health of a member's network interface is very important to the Cluster Alias (CLUA) subsystem as well as Cluster Application Availability (CAA) resource monitoring.

12.5.1 NIFF and the CLUA

When a cluster member enters run level 3, the aliasd(8) daemon is started; it in turn creates a child process known as aliasd_niff(8). The aliasd_niff process subscribes to all NIFF events from EVM although it is really only interested in those events that signal that a network interface has been marked as up (sys.unix.hw.net.niff.up) or down (sys.unix.hw.net.niff.down). When aliasd_niff receives one of these events from EVM, it notifies aliasd so that any necessary routing changes for the cluster aliases can be addressed. For more information on NIFF, see chapter 9 as well as the niff(7), nifftmt(7), and niffd(8) reference pages. For additional information on cluster aliasing, see chapter 16 as well as the aliasd(8)/aliasd_niff(8) reference page.

12.5.2 NIFF and CAA

NIFF also plays a role in CAA as well. If there are any network resources registered with CAA, then the resource manager (caad(8)) loads a monitor (network.so) that subscribes to the sys.unix.hw.net.niff.up and sys.unix.hw.net.niff.down events so that CAA will know when a resource needs to be relocated due to a network interface failure. We will be covering CAA in much greater detail in chapters 23 and 24. Additional information regarding CAA can also be found in the caa(4) and caad(8) reference pages as well as the TruCluster Server Cluster Highly Available Applications Guide.