16.15.1 When a node failure occurs on the master node
- Organization of this subsection
(1) System processing when a node failure occurs on the master node
When a node failure occurs on the master node, one of the slave nodes switches over to the master node, and the master node where the failure occurred is cut off from the multi-node configuration. The following figure shows the system processing when a node failure occurs on the master node:
|
|
![[Figure]](GRAPHICS/F0000934.GIF)
- Explanation
-
When a node failure occurs on the master node hadb01, one of the slave nodes switches over to the master node, and the node where the failure occurred is cut off from the multi-node configuration. Processing then continues on the master node hadb02 and the slave node hadb03.
In HADB, the HA Monitor multi-standby function is used to switch over the master node. The multi-standby function refers to the function for preparing multiple standby systems (slave nodes) for one active system (master node). The order of priority for slave nodes (which slave node becomes the master node) is set by using the initial and standbypri operands in the HA Monitor servers file. For details, see Managing servers and hosts when using the multi-standby function in the manual HA Monitor for Linux(R) (x86).
(2) Steps taken by the HADB administrator when a node failure occurs
To identify the cause of the node failure, take the action as described in 14.1 Error-handling flow. Perform the relevant procedure in the preceding reference, on the node where the node failure occurred.
Alternatively, obtain the following troubleshooting information on the node where the failure occurred:
-
HADB server troubleshooting information obtained by using the adbinfoget command
-
HA Monitor troubleshooting information obtained by using the HA Monitor monts command
After identifying the cause of the failure, you can return any nodes that were separated from the multi-node configuration, back to the configuration. For details about the procedure for returning nodes to a multi-node configuration, see 16.15.3 Returning a node to the multi-node configuration.
Note that nodes are always returned to a multi-node configuration as slave nodes.
(3) Notes
When there is only one master node in a multi-node configuration, and a node failure occurs on that master node, there is no master node to switch over to, so the multi-node configuration's HADB server terminates abnormally. If this happens, execute the HA Monitor monshow command on the master node where the node failure occurred, and check the status of the local host server.
If the local host server is in the ONL status (currently executing a process), execute the HA Monitor monend command, and terminate HA Monitor.