IBM SG24-5131-00 User Manual

Cluster Hardware and Software Preparation 

  • Scan the /tmp/hacmp.out file to confirm that the /etc/rc.net script has run 

successfully. Look for a zero exit status.

  • If IP address takeover is enabled, confirm that the /etc/rc.net script has 

run and that the service adapter is on its service address and not on its 
boot address.

  • Use the 

lssrc -g tcpip

 command to make sure that the inetd daemon is 

running.

  • Use the 

lssrc -g portmap

 command to make sure that the portmapper 

daemon is running.

  • Use the 

arp

 command to make sure that the cluster nodes are not using 

the same IP or hardware address.

Currently three types of non-TCP/IP networks are supported:

  • Serial (RS232) 

  • Target-mode SCSI

  • Target-mode SSA

While we use the word serial here to refer to RS232 only, in HACMP 
definitions, a “serial” network means a non-TCP/IP network of any kind. 
Therefore, when we are talking about HACMP network definitions, a serial 
network could also be a target-mode SCSI or target-mode SSA network. 

The following describes some cabling issues on each type of non-TCP/IP 
network, how they are to be configured, and how you can test if they are 
operational.

Cabling a serial connection requires a null-modem cable. As often 
cluster nodes are further apart than 60 m (181 ft.), sometimes 
modem eliminators or converters to fiber channel are used. 

If your cluster uses SCSI disks as shared devices, you can use 
that line for TMSCSI as well. TMSCSI requires Differential SCSI 
adapters (see Chapter 2.3.2.3, “Supported SCSI Adapters” on 
page 26). Because the SCSI bus has to be terminated on both 
ends, and not anywhere else in between, resistors on the 
adapters should be removed, and cabling should be done as 
shown in Figure 11 on page 77, that is, with Y-cables that are 
terminated at one end connected to the adapters where the other 
end connects to the shared disk device.