Serviceguard Network Manager: Inbound Failure Detection, March 2007

Introduction

This white paper provides a brief description about the default behavior of the Serviceguard network

failure detection mechanism. It also describes the network failure detection mechanism that deals with

inbound-only failures. It further explains the conditions under which the INONLY setting should be

applied.

Serviceguard Network Manager—default network failure

detection mechanism

The network failure detection scheme in Serviceguard Network Manager includes network driver error

handling and a polling mechanism. While performing periodic checks on cluster network interfaces,

Serviceguard Network Manager assesses the following to detect network interface failures:

• Network driver error notification

• Traffic statistics

Driver error notification

Whenever the network driver sends an error message indicating that a NIC has failed, Serviceguard

immediately declares that the NIC is bad. As long as there is a functioning standby network interface,

Serviceguard performs a failover. For example, when a NIC fails to send due to a severed link, the

network driver will notify Serviceguard, which in turn immediately declares the NIC failed upon

receiving this error notification.

Traffic statistics and polling mechanism

The purpose of Serviceguard Network Manager is to monitor the health of cluster member NICs. A

component of Serviceguard Network Manager is the polling mechanism. Serviceguard Network

Manager insures that both inbound and outbound messages increase and decrease for all network

interfaces in each active cluster member node. In addition to other network traffic, Serviceguard

Network Manager generates its own polling messages during every network polling interval. If there

is more than one network interface in a cluster member node, this mechanism performs local polling

(for example, polling among local network interfaces). If there is only one network interface in a

cluster member node, the mechanism will perform remote polling (for example, polling remote

network interfaces). The polling traffic mechanism ensures that there is a steady stream of traffic for

Serviceguard Network Manager to rely upon to determine if there is a problem with the LAN

interfaces.

In Figure 1, polling occurs between lan1 and lan0 on each active cluster member node.