While configuring VCS 4.0MP1 on a 3-node Solaris 8 cluster (actually, fsrac but this is irrelevant), I made a stupid mistake: copied the same /etc/llttab file to all three nodes and rebooted them. As a result, all three had the same node name, and, naturally, only one of them came up.
After realising my mistake, I expected to find the other two machines at ok> prompt. However, console access did not work, so the servers had to be power-cycled. I assume they did try to boot, then VCS (or llt?) detected node conflict, and executed `init 5' or some other command that results in powering down (on reasonably new Sun hardware anyway).
My problem with this is, if a server resides in a remote datacenter, someone needs to be there physically to bring the server back to life. Leaving the server at ok> prompt (`init 0' instead of `init 5') would be sufficient from VCS protection perspective, and would give a system administrator the vital option to work remotely.
Can this be done? VCS 4.0MP2? VCS 4.1?
Thanks,
Andy