IBM SG24-5131-00 사용자 설명서

Cluster Customization 

For example, a file system cannot be unmounted, because of a process 
running on it. Then, you might want to kill that process first, before 
unmounting the file system, in order to get the event script done. Now, since 
the event script didn’t succeed in its first run, the 

Retry feature enables 

HACMP for AIX to retry it until it finally succeeds, or the retry count is 
reached.

You must declare a shell (for example 

#!/bin/sh

) at the beginning of each 

script executed by the notify, recovery, and pre- or post-event processing 
commands.

Notify, recovery, and pre- and post-event processing do not occur when the 
force option of the 

node_down

 event is specified.

Synchronizing the cluster configuration does not propagate the actual new or 
changed scripts; you must add these to each node manually. Also, it is 
allowed to have different contents in these scripts on different nodes in order 
to be able to act upon different environments. However, the name of these 
scripts, their location in the file system, and their permission bits have to be 
identical.

To test the effect of running an event on your cluster, HACMP for AIX 
provides a utility to run an emulation of an event. This emulation lets you 
predict a cluster's reaction to an event as though the event actually occurred. 
The emulation runs on all active nodes in your cluster, and the output is 
stored in an output file. You can select the path and name of this output file 
using the 

EMU_OUTPUT

 environment variable, or, use the default 

/tmp/emuhacmp.out file on the node that invoked the Event Emulator.

For more information on event emulation, see these chapters: “Administrative 
Facilities” in the 

HACMP for AIX, Version 4.3: Concepts and Facilities, 

SC23-4276,

 and “Monitoring an HACMP Cluster” in the HACMP for AIX, 

Version 4.3: Administration Guide, SC23-4279.

The AIX Error Notification facility detects errors matching predefined 
selection criteria and responds in a programmed way. The facility provides a 
wide range of criteria that you can use to define an error condition. These 
errors are called 

notification objects.