I've a situation that has burned us several times in the past two weeks --
we have a two-node HACMP cluster (cascading) running SAP/R3. We had to do
a re-install of the primary node a month ago because of a rootvg disk
failure (we aren't mirroring rootvg -- this is one of the failures we have
HACMP for). Ever since, the values for hostname and uname -n have been
wrong. bear with me ---
en0 sp1-n25-en0 192.168.224.25 reliable ethernet
en1 colorado-bt 10.8.224.125 primary interface boot
en1 colorado 10.8.224.25 primary interface run
en2 colorado-stby 10.10.224.25 fallover interface
The requirement - uname -n and hostname MUST *always* return 'colorado';
we're getting 'sp1-n25-en0' for both during startup, with the hostname
switching to colorado somewhere in rc.net; uname -n never gets changed.
The challange, of course, is that the hostname/uname we want are not
values that can be entered in the SDR (en1 MUST come up on colorado-bt as
an IP address -- we got hosed in a fallover because the node did a
check-stop reboot and re-broadcast the run-time address *after* HACMP did
the address takeover and arp flush on the other node). What part of the SP
scripts do I need to nuke to use a hostname/uname NOT in the SDR?
Note that we did our re-install from a mksysb of this node; the values
were correct until the pssp software muked it up for us.