Electing The Node - Extreme Networks ExtremeWare XOS Guide Manual

Table 17: System redundancy terms (continued)
Term
Node Manager
Device Manager
Node Election
Checkpointing
The following sections describe the process, configuration, and monitoring of system redundancy.

Electing the Node

The node election process is a connectionless, un-reliable messaging mechanism between the nodes
installed in the chassis. The health of the node is also important in selecting the primary node. The
Device Manager collects the node health information and forwards that information to the Node
Manager. The Node Manager then computes the quality of the node which is later used in leader
election.
When two nodes exchange their health information, they come to a conclusion as to which is the
healthier node. Based on the election results obtained from all of the nodes, the healthiest node wins the
election criteria.
At the end of the election process, a primary node is selected along with a backup or secondary node.
The primary node runs the switch management functions, and the backup node is available to run the
switch management functions if the primary fails.
The parameters used to determine the primary node are:
• Node state—The node state must be ONLINE to participate in leader election and to be selected
primary. If the node is in the INIT, OFFLINE, or FAIL states, the node will not participate in leader
election. For more information about the node states, see "Viewing Node Statistics" on page 100.
• Configuration priority—User assigned priority. The configured priority is compared only after the
node meets the minimum thresholds in each category for it to be healthy.
• Control channel bandwidth—This is a function of the number of links available and the total
bandwidth of these links.
• Software health—This number represents the percent of processes available.
• Software version—Represents the software version the node is running.
• Health of secondary hardware components—Represents the health of the power supplies, fans, etc.
ExtremeWare XOS 10.1 Concepts Guide
Description
The Node Manager is a process that performs leader election between multiple nodes in the
system. In simple terms, the Node Manager elects the primary and backup MSMs. If there is
only one MSM installed, that MSM becomes the primary MSM.
To determine the primary node, the Node Manager sends hello messages to all of the nodes
in the system, computes the health of the system, and checks the parameters of each node
(for example the node state, configuration, priority, etc.).
The Device Manager is a process that runs on every node and is responsible for monitoring
and controlling all of the devices in the system. The Device Manager consists of a process
and a client library that is dynamically linked to every process that runs under XOS. The
library manages the communication of data between the device manager server and the
library so that all common system specific data required by this process is stored and
available to it immediate from it own private memory.
Node election is the actual process of electing the primary and backup node. Once elected,
the primary node coordinates all system activities between the nodes and becomes the main
computation and management point of contact. The backup node facilitates a faster recovery
if the primary node fails.
The process of copying the active state configurations from the primary MSM to the backup
MSM.
System Redundancy
97