The MultiNICB agent has two modes of operation, Base and Multipathing, which you can set with the UseMpathd attribute.
The value of the UseMpathd attribute is 0 by default for this mode. In Base mode, to monitor the interfaces that it controls, the agent:
The agent logs link failures and failovers when it uses either link- or probe-based detection.
If a NIC goes down, the MultiNICB agent notifies the IPMultiNICB agent. The IPMultiNICB agent fails over the virtual IP addresses to a different NIC on the same system. When the original NIC comes up, the agents fail back the virtual IP address.
Each NIC must have its own unique and exclusive base IP address, which the MultiNICB agent uses as the test IP address.
The MultiNICB agent, in Base mode, uses the following criteria to determine if an interface works:
The interface driver reports the status of the link. Note that not all drivers support this feature. Set the value of IgnoreLinkStatus to 1 to disable this test.
Set the LinkTestRatio attribute to a value greater than 0 to send ICMP echo request packets to a specified network host. You specify the network hosts in the NetworkHosts attribute. You must assign test IP addresses to the interface for probe-based detection. The test IP address is needed to send the ICMP packets, which determines the link's status. If you set the value of the LinkTestRatio attribute to 0, you do not need to assign test IP addresses.
If you specify no hosts in the NetworkHosts attribute, the agent uses the ICMP broadcast when the value of the NoBroadcast attribute is 0. It caches the sender of the first reply for future use as a network host. While the agent sends and receives ICMP packets, the IP layer is completely bypassed.
You can assign addresses and still do only link-based detection by setting the values of the LinkTestRatio and IgnoreLinkStatus attributes to 0. You can skip link-based detection (link driver tests) and only do ICMP tests if:
The MultiNICB agent performs both link-based detection and probe-based detection if:
The MultiNICB agent writes the status of each interface to an export information file, which other agents (like IPMultiNICB) or commands (like haipswitch
) can read.
During an interface failure, the MultiNICB agent fails over all logical IP addresses to a working interface under the same resource. The agent remembers the first physical interface from which an IP address was failed over. This physical interface becomes the "original" interface for the particular logical IP address. When the original interface is repaired, the logical IP address fails back to it.
To activate this mode set the value of the UseMpathd attribute to 1. The MultiNICB agent, in Multipathing mode, monitors Sun's IP Multipathing daemon (mpathd). The MultiNICB agent specifically monitors the failed flag on physical interfaces and the mpathd process. See the man page: in.mpathd (1M) for more information on this daemon.
Sun's mpathd daemon monitors the interfaces that are part of the IPMP group. The daemon:
The mpathd daemon can perform both link- and probe-based detection when test IP addresses are assigned to NIC interfaces.
The MultiNICB agent logs errors when the daemon is not running, or if a configuration path error exits. The mpathd daemon logs link failures and IP address failovers in the system log.
MultiNICB monitor agent function calls a VCS trigger in case of an interface going up or down. The agent passes the following arguments to the script:
The agent also sends a notification (which may be received via SNMP or SMTP) to indicate that status of an interface changed. The notification is sent using "health of a cluster resource declined" and "health of a cluster resource improved" traps. These traps are mentioned in the Veritas Cluster Server User's Guide. A sample mnicb_postchange trigger is provided with the agent. You can customize this sample script as needed or write one from scratch.
The sample script does the following:
MultiNICB agent Res. Name: Device en0 status changed from Down to Up.
The sample script does the following: