Re: Solaris8 to Solaris10... Cannot boot after Live Upgrade

On Feb 9, 9:47 pm, Randy Jones <r...@xxxxxxxxxxxxx> wrote:
Noel MiltonVegawrote:
Hi All:

Given:

- Sun v480
- Upgrade from Solaris 8 to Solaris 10 06/06 (upgraded via live
upgrade) SPARC
- Veritas VxVM & VxFS (4.x)
- Dual HBA's (vxdmpadm)

After splitting the root disk mirror, the live upgrade was performed.
(Solaris 8 is now on one disk; Solaris 10 is on the other disk).

The Solaris 10 disk has no knowledge of Veritas (for example, in /etc/
system
all Volume Manager and FS related entries were removed, and in general
all
steps were taken to make Veritas unknown to the Solaris 10 O/S). That
is,
after the live upgrade was performed (but before the reboot), pkgrm of
all
VRTS packages was performed.

After the upgrade completed, and the system was rebooted to boot the
Solaris 10 disk, the system went directly to Single User mode because
it said
it could not fsck the boot disk. We logged into the system, and did a
format(1M) which returned nothing (so we could not see the disks).
Fine.

So we booted to single user mode off the Solaris 10 CD-ROM and did
the following:

=============================
Step 1) boot single user mode off the Solaris 10 O/S CD
Step 2) mount -F ufs -o rw /dev/dsk/c1t0d0s0 /a
Step 3) (see rest that follows)...

rm -f /a/dev/dsk/*
rm -f /a/dev/rdsk/*
rm -f /a/etc/path_to_inst

You probably need to also remove /a/etc/path_to_inst.old
Plus then run devfsadm. My notes to myself on how to clone
a disk drive for a Blade 1000/2000:

Blade 1000/2000 disk drive clone notes
After cloning the harddrive do the following:
install cloned hard drive
boot from net or cdrom
mount root file system on /a
mv /a/etc/path_to_inst /a/etc/path_to_inst.original
rm /a/etc/path_to_inst.old
devfsadm -v -C -r /a
Then try to boot. If boot fails then take note as to what
file system can't be fsck'd, boot net or cdrom, mount disk on /a,
and edit /a/etc/vfstab to point to whatever location the
fsck was complaining about.
The diskdrives on the Blade 1000/2000 seem to magically change
from /dev/dsk/c0t1d0s0 to /dev/dsk/c1t0d0s0 to /dev/dsk/c0t0d0s0
to whatever...



(I never "rm" the /devices tree; I only append to it)

cp /tmp/root/etc/path_to_inst /a/etc/path_to_inst
cd /dev/dsk; find . | cpio -pudm /a/dev/dsk
cd /dev/rdsk; find . | cpio -pudm /a/dev/rdsk
cd /devices; find . | cpio -pudm /a/devices

/a/etc/vfstab checked for correct entries
=============================
Note: we also did it the devfsadm(1M) way.

Before rebooting (and actually before unmounting /a) we could see that
the resulting devices in /dev/(r)dsk resolved correctly to their
corresponding
entries in /devices (they were all there).

However when we cleanly unmounted /a and rebooted again to the
Solaris 10 O/S disk, the same thing happened -- i.e. it went to single
user mode because fsck could not stat(2) the bootdisk.

So we logged into the system (now in single user mode) and when
we checked, the links that the /dev/(r)dsk entries pointed to
were mysteriously gone. In other words everything in /dev/(r)dsk
was in tact, but what they pointed to in /devices were gone.

This is the case whether we did a reconfiguration reboot or a
non-reconfiguration reboot after doing the procedure above.

Why? Does anyone have an ideas as to whats happening?

TIA,
Noelle

--
----------------------------------
Randy Jones
E-Mail: r...@xxxxxxxxxxxxx
----------------------------------

Hi All...
Thank you for the advice. In fact we had tried the steps suggested
above already, but to no avail. Ultimately we rebuilt Solaris 10
using the same Live Upgrade procedure and the second time around, for
whatever reason, it worked.

I'd like to thank you guys for your responses.

Best Regards,
Noelle Milton Vega

.

Relevant Pages

  • SUMMARY: Disk contention
    ... I'm still convinced that it's not significant for big blocks (say 1 Mb I/O). ... S> due to the disk part. ... You can either use AT&T'ssfio library, or use Solaris 64-bit ... A significant bottleneck threshold occurs if b (processes ...
    (SunManagers)
  • Re: sparc 10 hanging
    ... could mount the disk and edit a file to reset the time ... Which is apparently updated every time anything is written to the boot ... So, since they are so similar between SysV (Solaris) and BSD, I ... would expect the entry in the equivalent file for SunOs 4.1.3 (a BSD ...
    (comp.sys.sun.hardware)
  • RE: ES40s wont boot!
    ... Subject: ES40's won't boot! ... a copy of the AS4100's system disk, ... they upgraded another ES40 at another ... site (in-box upgrade of CPUs from 667MHz to 833MHz), ...
    (comp.os.vms)
  • Re: Is Solaris 10 a good home os?
    ... Ok, then to HAVE lu-mirrored system disk before, but you still need to ... workstations using Live Upgrade. ... What's wrong with patches? ... It worked on 5 Solaris systems (4 ...
    (comp.unix.solaris)
  • Re: "no active partition"error (Solaris 10/x86)
    ... I just installed Solaris 10 on a host with an LSI 320-2 RAID ... Is this the only disk in the box? ... You might have a corrupt boot sector. ... Boot Sector and then do the install again. ...
    (comp.unix.solaris)