ACC Installation and Configuration Guide

96 AppendixC
ACC Service Guard Configuration Notes
Overview
Overview
In the examples, there are two systems that each have one or more ACC
cards, using the dual connect ACC distribution panel accessories. One
system is a primary, and one is a secondary. The primary system has the
ACC links in active use, and the secondary either does not start the ACC
subsystem, or has each link (pmode) in the loopback mode.
There are 2 cases that need to be considered.
• In the first case, the primary system has an unexpected, catastrophic
failure or HP-UX panic.
• In the second case, the primary system is shutdown (gracefully).
The SG cluster configuration is controlled by a script, run on each system
when the system joins the cluster. The script is i nvoked with options that
indicate which subsystems are to be made active, and which are to be
inactive.
Tocontrolthebehavior,inbothcasesyouneedtodothefollowing:
1. In the networkconfiguration file,/ o pt/acc/cfg/loopback.answ (orthe
filename you choose), define the port mode for each link to be LBACK.
YoualsoneedtodefineTerm’sformonitor(oneforeachACCcard).
2. In the control file, /opt/acc/cfg/zmasterd_list, add statements to
start watch and stop watch .Thewatch daemon is the heart beat
generation for the ACC monitor protocol.
When started, watch enables monitor, and then sends a heartbeat to
monitor. Monitor will reset the ACC, if a heatbeat is missed. The ACC
reset will put the ACC ports in LBACK (tri-state).
This handles the case of an unexpected failure where the Service
Guard cluster control script does not run with the stop option.
Youalsoneedtomakethefilepathforzmon correct
(/opt/acc/cfg/loopback.tmem).
3. In the start up, /etc/rc.config .d/acc file, make sure that the start flag
is 0 (START_ZCOM=0). This allows the Service Guard script to
control when to s ta rt/stop ACC, so that ACC will not restart
automatically following a panic, or powerout event and interfere with
the other system.