2.19.3 Managing nodes with HA Monitor
HA Monitor is required to use the multi-node function. HA Monitor is a program product that monitors node statuses and separates any node on which a failure occurs.
This subsection explains the roles (functions) of HA Monitor in the multi-node function. For details about HA Monitor, see the manual HA Monitor for Linux(R) (x86).
- Organization of this subsection
(1) Determining node types
HA Monitor determines which node is to be used as the master node, based on the specification in the servers file for defining the HA Monitor environment settings. The specification of the initial operand for each node in the servers file governs the determination. If online is specified for that operand, the node becomes the master node (active system of HA Monitor). If standby is specified for that operand, the node becomes a slave node (standby system of HA Monitor).
The following figure shows the relationship among the master node, slave nodes, and the active system and standby system of HA Monitor.
|
|
![[Figure]](GRAPHICS/F0000105.GIF)
When the HADB servers in a multi-node configuration are started, the master node and slave nodes are determined according to the specification in the initial operand.
- Note
-
To start HADB servers in a multi-node configuration, execute the adbstart command, and then execute HA Monitor's monbegin command. During this process, you must execute the monbegin command before processing of the adbstart command is completed.
(2) Monitoring for failures
HA Monitor monitors for the failures described in the table below. When any of the following failures is detected, the node on which the failure occurred is separated from the multi-node configuration.
|
No. |
Failures monitored by HA Monitor |
|---|---|
|
1 |
Hardware failure |
|
2 |
Hardware power failure |
|
3 |
Kernel failure |
|
4 |
HA Monitor abnormal termination |
|
5 |
HADB server abnormal termination (server process abnormal termination) |
The failures listed in the above table are called node failures.
- ■ Key points when reading the manual HA Monitor for Linux(R) (x86)
-
The failures monitored by HA Monitor are host failures (failures 1 through 4 in the above table) and server failures (failure 5 in the above table). Since the multi-node function works according to HA Monitor's monitor mode, you must create a server-monitoring command if you want to detect server failures. For details about how to create a server-monitoring command, see (e) Creating a server-monitoring command in (4) Creating environment variable definitions for commands, and creating commands under 16.3.4 Setting up an HA Monitor environment.
(3) Separating a node when a failure occurs
If a node failure occurs as described in (2) Monitoring for failures, HA Monitor functionality is used to separate the node on which the failure occurred from the multi-node configuration. This allows operations to continue by using the remaining nodes.
The node separation performed in the event of a node failure is managed by the HA Monitor. For details, see 2.19.4 Separating a node from a multi-node configuration.