I'm attempting to upgrade a couple of servers (E4500) from 2.6 to 8, and the
approach I'm trying to take is to copy the root file system from a working
server to an empty disk on the other servers.
The Solaris 8 disk image on the source server is not the active boot device,
i.e. it is static, and has been highly customized in terms of add-on
products and tweaks. It was initially built via jumpstart.
This approach was also taken in an attempt to save time, by minimizing the
downtime of jumpstarting, and fully patching, and installing add-on
packages, etc. In theory, the disk image being copied is almost ready for
booting in our production environment.
I've not done any jumpstart scripting and thought this approach would be
quicker in the short term, boy was I wrong.
I've learned a lot about hardware devices and the importance of /dev/dsk,
/dev/rdsk, /etc/path_to_inst, /devices, etc...
However, I need to proceed with this approach due to timing issues and fast
A server was upgraded yesterday and is now running Solaris 8. However, the
disk controller numbers went from c0/c1/c2 on 2.6 to c9/c12/c13 on 8. It's
running ok, but I'm concerned and puzzled by this.
The source disk image has had the /dev/dsk and /dev/rdsk directories
emptied. After the copy (via ufsdump/ufsrestore) the 2.6 version of
path_to_inst was restored to the target disk.
After the copy, the target server was booted with a Solaris 8 cd, and
drvconfig, devlinks, disks, tapes, executed to build the device entries.
Then a reconfiguration boot was done.
Admitting this might not be the best approach, what should be done to get it
to work better.
Should the target path_to_inst be empty, in addition to the /dev/dsk,
/dev/rdsk, and /devices?