Service Guard

[Rob Saunders]

I seem to recall that when clustering machines, the systems' main boards must have different identities to avoid conflicts. This used to be done with a DIP switch setting on the main board but is now done thro' the PDC Service menu.
My question is; Can you determine these settings from multi-user mode (as root)? I've tried ioscan and its variations but all seem to show the same values, which I know is mistaked ('cos the system is working okay).

[James R. Ferguson · Accepted solution]

Hi Rob:

'stm' for the processor(s) might give you the information you seek.

# echo "selclass qualifier processor;info;wait;infolog"|cstm > /tmp/procinfo

...JRF...



...JRF...

[Pedro Sousa]

Once more, James is totally correct.

If you don't have Diagnostics installed you can get it at http://www.software.hp.com/cgi-bin/swdepot_parser.cgi/cgi/displayProductInfo.pl?productNumber=B6191AAE
good luck.

[Stephen Doud]

The CSTM command didn't work for me (and CSTM can be CPU-intensive)

Try this simple command:

# ioscan | grep -e bus -e ctl
0/0/1/0 ext_bus SCSI C895 Fast Wide Single-Ended
0/0/1/0.7.0 ctl Initiator
0/0/2/0 ext_bus SCSI C875 Ultra Wide Single-Ended
0/0/2/0.7.0 ctl Initiator
0/0/2/1 ext_bus SCSI C875 Ultra Wide Single-Ended
0/0/2/1.7.0 ctl Initiator
0/8/0/0 ext_bus PCI Ethernet (10110019)
0/12/0/0 ext_bus SCSI C875 Fast Wide Differential
0/12/0/0.7.0 ctl Initiator
0/12/0/1 ext_bus SCSI C875 Fast Wide Differential
0/12/0/1.7.0 ctl Initiator

If the ext_bus is SCSI, check the next line for a ctl - the SCSI adapter card. The hardware path is on the left.

Example: The last adapter listed has an address of 0/12/0/1.7.0 The actual SCSI address is 7.

[James R. Ferguson]

Hi Stephen:

I agree with you regarding SCSI addresses via ioscan. This is the obvious "point of interest" from the cluster's disk perspective.

The reference to "main boards" caused me to think of 'stm' and I wasn't exactly sure what Rob was interested in ascertaining. The command I offered should also have used 'cpu' and not 'processor' as the qualifier. I was at home and working from memory.

Thanks for following up.

...JRF...

[Carsten Krege]

The output that James creates with his command (I was pretty impressed!)( also lists the "software id" of the machine, also printed by the command

# uname -i

This software id must be unique for any machine in a SG cluster(in fact for any machine!), otherwise you would definitely face problems with SG. This should never happen, but if it does, the software id should only changed by HP personnel.

Carsten