3.3.1 Preventing concurrent host resets
Concurrent host resets means that the active and standby systems both perform host reset at the same time. HA Monitor prevents concurrent host resets from occurring, because if concurrent host resets do occur, both hosts are terminated and job processing is suspended. This subsection explains how to prevent concurrent host resets.
- Organization of this subsection
(1) Configurations in which concurrent host resets might occur
Depending on the hot-standby configuration and the functions in use, the active and standby systems might both perform reset at the same time, resulting in termination of jobs.
Especially in the hot-standby configurations described below, the active and standby systems might both be reset at the same time. Therefore, you must determine which host is to be reset first. If both hosts detect host failures at the same time, the host that issues a reset request first is called the host that has reset priority.
-
Mutual hot-standby configuration
The reset priority must be determined because both hosts are active systems and constitute each other's standby systems. If the system being used is BladeSymphony and SVP supports the cluster management function, you can also choose a method for preventing concurrent host resets without having to determine the reset priority. For details about the cluster management function, see the hardware-related documentation.
-
Resetting the standby system in a 1-to-1 switchover configuration
If the standby system is to be reset in the event of a failure in the standby system, both hosts might issue a reset at the same time. Therefore, you must determine the reset priority beforehand. For details about standby system reset, see (2) Resetting the standby system.
The following figure shows a configuration in which concurrent host resets might occur.
|
In the configuration shown in the figure above, if two hosts detect the host failure at the same time, both hosts issue a reset request at the same time. As a result, both hosts might be reset simultaneously.
(2) Required environment settings
You must determine the host that is to have reset priority in order to prevent concurrent host resets. HA Monitor identifies the host that has reset priority on the basis of the information specified in the cpudown operand in the HA Monitor environment settings. By default, a reset request from the active system has priority over a reset request from a standby system.
If you want to prevent concurrent host resets without having to determine if the host has reset priority, specify system in the cpudown operand in the HA Monitor environment settings. This specification is applicable only when a mutual hot-standby configuration is used, the system being used is BladeSymphony, and the SVP supports the cluster management function. If VMware ESXi-based virtualization is used, system cannot be specified.
For details about the HA Monitor processing for preventing concurrent host resets, see 4.2.2 Hot standby operation when both hosts detect a failure at the same time.