2.6 to 8 upgrade, copy boot disk, device issues

[Rob Gates]

Have you ever looked at the Sun package Live Upgrade ? I have used it to
upgarde a couple of our servers and it went very well with limited
disturbance to the systems.
"Peter Vines" <vines@nospam.virginia.edu> wrote in message
news:beu8nt$ni5$1@murdoch.acc.Virginia.EDU...

> 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

> approaching deadlines.
>
> 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?
>
>

[Darren Dunham]

Atif Munir <atifch@webcarriers.com> wrote:

> Hello,

> Just boot via cd room and mount / slice
> touch /reconfigure
> and init 6
> it will make /dev and /devices again and your os will boot.

Uh, no.

Both a boot -r or a boot with /reconfigure don't do anything useful
until after the root filesystem is mounted read-write. If the devices
are screwed up, that won't happen.

Boot -r can fix *some* devices, but not those associated with the root
disk. Those must be fixed externally before the boot will succeed.

--
Darren Dunham [email]ddunham@taos.com[/email]
Unix System Administrator Taos - The SysAdmin Company
Got some Dr Pepper? San Francisco, CA bay area
< This line left intentionally blank to confuse you. >

[sharona]

"Peter Vines" <vines@nospam.virginia.edu> wrote in message news:<beu8nt$ni5$1@murdoch.acc.Virginia.EDU>...

> 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
> approaching deadlines.
>
> 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?

Gee, if it were me I would upgrade the machine though jumpstart. you
can edit the profile and preserve partitions so you don't lose data.
There are so many differences between 2.6 and 2.8 that I wouldn't even
attempt coping from one to the other. Jumpstart would work a whole
lot faster. I usually set up the jumpstart and kick it off right
before I leave. When I come back in the next morning, everying is
done!

Sharona