N.4 Terminating monitoring during scale in
If an agent host is deleted during scale in, the health check function detects that the agent host has stopped or that an error has occurred, and then performs the automated action for the JP1 event that is registered in the manager host. As a result, the agent host is deleted from the system's monitoring targets.
- Organization of this subsection
(1) Detecting a scale in by using the health check function
Scale-in events are detected by the health check function (monitoring of remote hosts).
-
When the JP1/Base service stops, an agent sends the manager an event notifying the manager that the service has stopped (event ID: 4725). For details about this event, see 17.3.1(42) Details about event ID 00004725.
-
If, for some reason, the event notifying the manager that the service has stopped cannot be received, an agent error is detected and an error detection event (event ID: 4749 or 474D) is issued.
-
If an error was detected during normal monitoring, an error detection event (event ID: 4749) is issued. For details about this event, see 17.3.1(49) Details about event ID 00004749.
-
If the stopping of the service cannot be distinguished from an error, an error detection event (event ID: 474D) is issued. For details about this event, see 17.3.1(53) Details about event ID 0000474D.
-
(2) Clearing the network settings and the monitoring settings
The network settings and monitoring settings are cleared by the automated action for the health check stop notification event (event ID: 4725) and by the automated action for the error detection event (event ID: 4749 or 474D), as follows.
-
Information about the agent host that was scaled in is deleted from the manager host's hosts file.
-
The agent host is deleted from the health check definition file (jbshc.conf) and then JP1/Base is reloaded to apply the definition.
-
The agent host is deleted from the configuration definition file (jbs_route.conf).
If the value of the JbsrtDistribByScaleIn setting parameter is YES in the jbsas_setup_manager.ini file, the configuration definition is distributed. If the value is NO, the configuration definition is not distributed. It is distributed the next time a scale out occurs. The initial value is NO.
-
If JP1/PFM is used for monitoring, the agent's monitoring information is deleted.
If a JP1 event is reported due to a temporary error, but a scale out has not occurred, these settings are not cleared even when an automated action is executed.
(3) Checking whether monitoring settings were cleared successfully
You can check whether monitoring settings were cleared successfully by checking the termination code of the automated action for the stop notification event (event ID: 4725) or for the error detection event (event ID: 4749 or 474D). If the automated action terminated normally, the termination code will be 0. If the action terminated abnormally, the termination code will be a nonzero value.
If the automated action terminated abnormally, remove the cause of the error, and then re-execute the automated action.