SCO 5.0.7 Esxi 4.1 Backup Edge 2.3

[Steve M. Fabac, Jr.]

I worked with a client's Windows system administrator to
test migration of an SCO 5.0.7 system to his VMWare ESXi
platform.

The restoration of the Backup Edge DVD media backup to the
virtual machine was partially automatic but required manual
steps as well.

We set the SCO virtual machine to boot the host DVD drive
before the hard disk and I booted with:

defbootstr Sdsk=wd(0,0,0,0) Srom=wd(1,0,0,0)

and Backup Edge RE2 on the backup media booted ok and tested
and loaded the RE2 files from the DVD.

I tried the "automatic" restore and it failed when trying to
access the DVD to restore the backed up files. I did not
record the exact error but is said something about getting the
wrong number of bytes during the read and locked up.

I rebooted the virtual machine and the RE2 media again
and used the utilities -> shell and ran divvy manually.
All file systems were created by the automatic restore attempt
except one that shows "unknown file type." I manually
created the missing file system and ran fsck on all the
file systems. Fsck finished without a problem.

I then tried the restore all files function of RE2. The status
screen that reports mounting the file systems announced that
lost+found already existed (was created by the "automatic"
restore attempt) but the restore failed with the wrong
byte count message and locked up.

I rebooted again, ran fsck on all file systems (in all the
fsck run, none reported errors), mounted all the
file systems as needed, changed to /mnt and executed
edge xvbf 64 /dev/cd0 and Edge restored the files.

When the restore was finished, I unmounted all the file systems
and shut down RE2 menu and rebooted using the same defbootstr.

The system came up and I edited /etc/default/boot and added
Sdsk=wd(0,0,0,0) and Srom=wd(1,0,0,0) to the default boot string.

Problem: When I used shutdown -y -g0 -i6 (to update the boot
file system /etc/default/boot), the system panicked with Trap E
when it came back up. I booted the RE2 media again, mounted
the boot file system and removed the Sdsk and Srom declarations
from the /etc/default/boot file. Upon rebooting, I used
defbootstr Sdsk=wd(0,0,0,0) Srom=wd(1,0,0,0) again and the system
panicked again. Power-off the VM and back on and then boot with
the defbootstr again and the system booted ok.

I edited /etc/conf/sdevice.d/aacraid to disable the SCSI raid
controller, and edited /etc/conf/cf.d/mscsi to change the hard
drive from aacraid Sdsk 0, 0, 0, 0 to wd Sdsk 0, 0, 0, 0 (the DVD drive
was already at wd 1, 0, 0, 0), relinked the kernel and rebooted.

The system booted ok without the defbootstr Sdsk=wd(0,0,0,0)
Srom=wd(1,0,0,0).

Reconfiguring the kernel by changing the network interface card
and rebuilding the kernel and reboot did not panic.

Here is the remaining problem: I can't get Backup Edge 2.3 to
backup to the host DVD drive. Trying to run edgemenu and
listing the contents of the dvd media used to restore the SCO
system to the ESXi system produces the same wrong byte count
error and locks up the console session. Pressing Alt-F4 or
alt-F6 switches console screens and allows another log in.

All efforts to reconfigure the DVD in the SCO VM properties
(IDE slave on primary controller, SCSI on LSI parallel, SCSI
on Buslogic) fails to produce a DVD configuration where
dtype /dev/cd0 says anything other then "no such device or
address."

Cdrecord -scanbus shows the device for the DVD but trying to
use cdrecord to write an ISO to the DVD-R media fails without
a defined error message (No message: wrong disk/no disk) just
exits without writing to the DVD-R media.

My only experience with VMWare prior to working with the client's
ESXi system was doing essentially the same thing on VMWare
player 3.1.3 on my XP home system where I have no trouble
specifying the hard disk as the master on the primary IDE
controller and the DVD drive on the blc (Buslogic) SCSI
controller:

*
* ha = device name of SCSI host adapter driver e.g. ad
* attach= device name of attached SCSI device driver e.g. Sdsk
* number= host adapter number
* ID = controller number
* lun = logical unit number
* bus = host adapter bus number
*
*ha attach number ID lun bus
*
blc Srom 0 0 0 0
(END)

Odd that I now look at it, no entry for the IDE hard disk?

/etc/default/boot
#ScoAdminInit BOOTMNT {RO RW NO} RO
#
DEFBOOTSTR=hd(40)unix swap=hd(41) dump=hd(41) root=hd(42) kbm.wheel=no
AUTOBOOT=YES
TIMEOUT=10
FSCKFIX=YES

# hwconfig -h | less
device address vec dma comment
======== ============= === === ================================================
kernel - - - rel=3.2v5.0.7 kid=2003-02-18
cpu - - - unit=1 family=15
cpuid - - - unit=1 vend=AuthenticAMD tfms=0:15:5:3
fpu - 13 - unit=1 type=80387-compatible
pci 0xcf8-0xcff - - am=1 sc=0 buses=35
PnP - - - nodes=0
clock - - - type=TSC/3.214426262Ghz
serial 0x3f8-0x3ff 4 - unit=0 type=Standard nports=1 base=0 16550A/16
serial 0x2f8-0x2ff 3 - unit=1 type=Standard nports=1 base=8 16550A/16
console - - - unit=vga type=0 num=12 scoansi=1 scroll=50
adapter 0x10c0-0x10c3 11 - type=blc ha=0 id=7 fts=stn
floppy 0x3f2-0x3f7 6 2 unit=0 type=135ds18
kbmouse 0x60-0x64 12 - type=Keyboard|PS/2 mouse id=0x00
udi - - - UDI environment
adapter - - - ha=0 type=usb_msto UDI SCSI HBA
pnt 0x2000-0x2012 10 - pnt0 Ver: 4.06 addr=00:0c:29:0c:ea:4f
cd-rom - - - type=S ha=0 id=0 lun=0 bus=0 ht=blc unit=0
disk 0x1f0-0x1f7 14 - type=W0/0 unit=0 cyls=520 hds=128 secs=63
Srom-0 - - - Vnd=ATAPI Prd=iHAS124 Y Rev=BL0V
(END


So ESXi running on a Dell system with 32G memory, SSD in RAID-1 emulating
IDE to the SCO system, and a DVD-RW slim-line drive is having trouble
with DVD access by the guest SCO 5.0.7 OS where VMWare player 3.1.3 has
no trouble accessing the WinXP host DVD drive.

The customer's system restored to the ESXi system is fast when running
disk intensive reports: Completes a report that takes 5 minutes on the
SCSI RAID-1 SCO hardware in only 37 seconds under ESXi.

SCO 5.0.7 system with MP5: 4G RAM, 2.4G 4 core Intel CPU Adapted
parallel SCSI RAID-1 10K disks.

Nagging problems that are preventing us from trying the ESXi hosted
system in production (besides the inability of making DVD-R backups using
Backup Edge) are repeated instances of kernel panics with Trap E on boot
up. No other panics have occurred in the short period of testing other
then the occasional boot up panic. Trying to mount the CD in the SCO
VM using the ESXi Vsphere console icon results in SCO VM becoming
unresponsive for up to three minutes.

Looking for suggestions from anyone who has used ESXi to host
SCO 5.0.7 (not SCO 5.0.7v) in evaluation or production and what
they find as necessary configuration parameters.

--
Steve Fabac
S.M. Fabac & Associates
816/765-1670

[Jean-Pierre Radley]

Steve:

Why aren't you a member of iXorg?

Check us out at ixorg.org.

--
JP

[BobD]

Hi Steve,

I have attached a pdf of a Support Note I wrote when I installed 5.0.7
under VMWare ESXi 4.1 (although I have also done it on V5). Rather
than restoring from a backup, I did a clean install of 5.0.7 onto VM5,
then installed/restored programs and data for my customer. This
Support Note refers to other Notes, but they are just standard
installation Notes, not relevant to the VM5 installation.

I have this operational in a live environment at a customer site, and
I also use it for my own Support System.

I don't use Backup Edge for backups, we have our own routine using
cpio. This seems to work ok with SCSI DDS Tape Drive at my Customer
site. I don't want to install a tape drive on my own VM Machine, so I
am still thinking about how to backup my machine using the removable
hard disk we now use for backup (hitting the maximum file size problem
when trying to us cpio. or tar). As we don't actively develop under
SCO Unix anymore, not too much of a problem, as I have backups of our
legacy system I can restore if necessary).

Hope the attached notes help

BobD

[Bela Lubkin]

Steve M. Fabac, Jr. wrote:

> defbootstr Sdsk=wd(0,0,0,0) Srom=wd(1,0,0,0)

"Sdsk=wd" is almost always (100%) wrong. OSR5 does not internally
convert ST506/IDE access to SCSI (Linux, and I think Solaris and BSD and
probably everything else that's been under Active Development for the
last 10 years does). All you are likely to accomplish with that is to
mess up the SCSI disk driver -- which probably isn't a problem in either
a physical system with only IDE disks, or a virtual system where all
your hard disks are probably the same virtual flavor. None of which,
physical or virtual, are *ever* "Sdsk=wd".

There have, in the fullness of time, existed a few devices which *did*
end up legitimately seen as "Sdsk=wd". Those are the very limited range
of devices which were ever available as *ATAPI* random access block
devices: LS-120/240, Clik, Zip, Jaz, Rev, Orb, Sparq, EZFlyer, and other
such third-rate removable storage devices. I know for sure that at
least the LS-120 and Zip flavors work(ed) with OSR5 and were operated
through Sdsk -> ATAPI -> wd.

Be that as it may. Nothing that you would feel solidly in favor of (A)
_calling_ a hard drive and (B) _using_ as a hard drive, has ever used
that path.

> Problem: When I used shutdown -y -g0 -i6 (to update the boot
> file system /etc/default/boot), the system panicked with Trap E
> when it came back up. I booted the RE2 media again, mounted
> the boot file system and removed the Sdsk and Srom declarations
> from the /etc/default/boot file. Upon rebooting, I used
> defbootstr Sdsk=wd(0,0,0,0) Srom=wd(1,0,0,0) again and the system
> panicked again. Power-off the VM and back on and then boot with
> the defbootstr again and the system booted ok.

The panic may have been a known and now fixed problem of OSR5 on VMware
environments. Some timing sequence would occasionally go wrong. I
forget if it was fixed in OSR507MP5 or only in "OSR507V2".

> All efforts to reconfigure the DVD in the SCO VM properties
> (IDE slave on primary controller, SCSI on LSI parallel, SCSI
> on Buslogic) fails to produce a DVD configuration where
> dtype /dev/cd0 says anything other then "no such device or
> address."

My guess is that you've just been fooled by ESX user interface. When
you map something (physical CD drive or ISO image) to virtual device
(ATAPI drive at IDE0:1, etc.) -- you are only establishing a mapping for
the host to remember. You aren't "putting a disc in the drive". To do
that you must additionally check at least one of "[x] Connect at power
on" or "[x] Connected now" (not exact wordings). The first means that
it will be connected *next* time the VM is powered on (and all
subsequent times until you change it). The second is greyed out unless
the VM's powered on; then it means "insert it right now" (but don't
remember for subsequent power-ons). I fool with CD/ISO media and ESX
many times a day and I still screw this up half the time...

> Cdrecord -scanbus shows the device for the DVD but trying to
> use cdrecord to write an ISO to the DVD-R media fails without
> a defined error message (No message: wrong disk/no disk) just
> exits without writing to the DVD-R media.

This is because ESX isn't presenting you a hot-plug drive. Whether
ATAPI or SCSI emulation, the OS sees a "drive" already present at
power-on. The two "connect" options are about whether a "disc" is
inserted into the virtual drive or not.

> My only experience with VMWare prior to working with the client's
> ESXi system was doing essentially the same thing on VMWare
> player 3.1.3 on my XP home system where I have no trouble
> specifying the hard disk as the master on the primary IDE
> controller and the DVD drive on the blc (Buslogic) SCSI
> controller:

VMware Hosted products (Workstation, Fusion, Player) have the same
distinction, but I think the UI is more sensible. When you "add" a
device and give it a mapping to a medium at the same time, it sensibly
decides you meant to "insert" that medium into the drive.

> *ha attach number ID lun bus
> *
> blc Srom 0 0 0 0
>

> Odd that I now look at it, no entry for the IDE hard disk?

Because it's an IDE hard drive. Not ATAPI.

> adapter 0x10c0-0x10c3 11 - type=blc ha=0 id=7 fts=stn

> adapter - - - ha=0 type=usb_msto UDI SCSI HBA

> cd-rom - - - type=S ha=0 id=0 lun=0 bus=0 ht=blc unit=0
> disk 0x1f0-0x1f7 14 - type=W0/0 unit=0 cyls=520 hds=128 secs=63
> Srom-0 - - - Vnd=ATAPI Prd=iHAS124 Y Rev=BL0V

This actually seems rather wrong, I'm not sure how you got here...

"adapter... blc" is a BusLogic BT946 PCI SCSI. "cd-rom... ht=blc" is a
CD-ROM drive on the BusLogic. But "Srom-0... Vnd=ATAPI" means you
actually accessed an ATAPI drive. Oh wait, maybe I see. This is a
physical ATAPI drive on the ESX host. ESX has laundered it through as
SCSI to the virtual BLC, stamping "ATAPI" into the vendor field.

...

Ok, ignoring all of the above, let me just make a few comments about the
state of OSR5-under-ESX these days.

1. There are few reasons to install a virtual SCSI adapter into an OSR5
VM. Use IDE virtual hard disk, IDE (ATAPI) virtual CD/DVD drives.

Possible uses for SCSI: (a) if you're migrating an existing SCSI-
based system into a VM and trying to map it as exactly as possible
(you'd still have to change the SCSI driver unless it's using lsil
adapters, but you could do a simple substitution in mscsi). (b) to
pass through a SCSI (or possibly USB) tape drive: not sure if this
would work, but if you wanted to try, you definitely want SCSI at the
virtual end of the chain. But you'd still want IDE for disk & CD.

2. OSR507 works better than OSR506(or earlier), mainly because when SCO
started doing virtualization-specific fixes, they were put into
OSR507 maintenance supplements. Fixes available only for 507
include: making the mouse work (http://kb.vmware.com/kb/1016914
or http://sco.com/ta/127469 -- space.c edit required);
making X work through multiple reboots and VM reconfigs
(http://kb.vmware.com/kb/1016912 or http://sco.com/ta/126555);
several panic fixes.

3. OSR507V works better than OSR507. I believe it's currently up to
"OSR507V3". As far as I could tell, 507V2 was a moderately
significant update with some V-specific fixes; 507V3 not so much.
Fixes only available in the "V" series include further panic fixes;
data corruption fixes for SCSI (problem didn't exist in IDE); the
smarts to not demand multiple licenses for a single multicore
CPU; and I forget what else (I know I'm forgetting at least one
important one). Also, 507V includes a subset of VMware Tools
(http://sco.com/ta/127360).

A supplement existed (hopefully still does) which delivered the 507V
fixes into a regular 507 install. That's the good news. The bad
news: it's intended to not work without a 507V license (as far as I
know / remember). Related bad news: SCO/unXis do not support any
releases other than 507V (and 600V) on VMware platforms; and VMware
support only 506, 507 and 600 (*NOT* anything with a 'V' suffix).

Ah, OSR507V3 is hidden here:

http://sco.com/support/update/download/product.php?pfid=16&prid=25

and most specifically here:

http://sco.com/support/update/download/release.php?rid=423

>Bela<

[Steve M. Fabac, Jr.]

Bela Lubkin wrote:
> Steve M. Fabac, Jr. wrote:
>
>> defbootstr Sdsk=wd(0,0,0,0) Srom=wd(1,0,0,0)
>
> "Sdsk=wd" is almost always (100%) wrong. OSR5 does not internally
> convert ST506/IDE access to SCSI (Linux, and I think Solaris and BSD and
> probably everything else that's been under Active Development for the
> last 10 years does). All you are likely to accomplish with that is to
> mess up the SCSI disk driver -- which probably isn't a problem in either
> a physical system with only IDE disks, or a virtual system where all
> your hard disks are probably the same virtual flavor. None of which,
> physical or virtual, are *ever* "Sdsk=wd".

I had never noticed that in mscsi before for IDE disk. In my defense
I was booting a restored system that has Adaptec aacraid as the root
disk controller (RAID1) for the hardware hosting the physical 5.0.7 system.

Had I not used the Sdsk=wd(0,0,0,0) in the defbootsr when booting the
Backup Edge RE2 image, I would have hit the condition "No root disk
controller."

I tried both options without success. Connect at power up still fails
the dtype /dev/cd0 test after booting the 5.0.7 VM and placing a
valid media (DVD disk with data on it) in the ESXi host machine DVD
drive. Clicking the CD icon in the vShpere console session for the SCO
VM offers many options to "connect" the drive. One item that passed
the dtype /dev/cd0 test was selecting "connect -> D:" and placing the
disk in the DVD drive on the Windows XP system where the Vsphere client
software is running. Clicking the CD icon and selecting "connect -> ESXi"
host drive with the DVD media in the drive takes a long time to execute
and then fails the dtype /dev/cd0 test.

Perhaps ESXi is not configured correctly? The Host device offered is:

/vmfs/devices/cdrom/mpx.vmhba36:C0:TO:L0

>
>> Cdrecord -scanbus shows the device for the DVD but trying to
>> use cdrecord to write an ISO to the DVD-R media fails without
>> a defined error message (No message: wrong disk/no disk) just
>> exits without writing to the DVD-R media.
>
> This is because ESX isn't presenting you a hot-plug drive. Whether
> ATAPI or SCSI emulation, the OS sees a "drive" already present at
> power-on. The two "connect" options are about whether a "disc" is
> inserted into the virtual drive or not.

The "connect" options did not work with DVD media with content or
with DVD-R blank media.

>
>> My only experience with VMWare prior to working with the client's
>> ESXi system was doing essentially the same thing on VMWare
>> player 3.1.3 on my XP home system where I have no trouble
>> specifying the hard disk as the master on the primary IDE
>> controller and the DVD drive on the blc (Buslogic) SCSI
>> controller:
>
> VMware Hosted products (Workstation, Fusion, Player) have the same
> distinction, but I think the UI is more sensible. When you "add" a
> device and give it a mapping to a medium at the same time, it sensibly
> decides you meant to "insert" that medium into the drive.
>
>> *ha attach number ID lun bus
>> *
>> blc Srom 0 0 0 0
>>
>> Odd that I now look at it, no entry for the IDE hard disk?
>
> Because it's an IDE hard drive. Not ATAPI.
>
>> adapter 0x10c0-0x10c3 11 - type=blc ha=0 id=7 fts=stn
>> adapter - - - ha=0 type=usb_msto UDI SCSI HBA
>> cd-rom - - - type=S ha=0 id=0 lun=0 bus=0 ht=blc unit=0
>> disk 0x1f0-0x1f7 14 - type=W0/0 unit=0 cyls=520 hds=128 secs=63
>> Srom-0 - - - Vnd=ATAPI Prd=iHAS124 Y Rev=BL0V
>
> This actually seems rather wrong, I'm not sure how you got here...

The above output of hwconfig -h was on VMWare Player 3.1.3 on my Windows
XP system. The dvd is Lite On iHAS124 SATA DVD RW drive.

>
> "adapter... blc" is a BusLogic BT946 PCI SCSI. "cd-rom... ht=blc" is a
> CD-ROM drive on the BusLogic. But "Srom-0... Vnd=ATAPI" means you
> actually accessed an ATAPI drive. Oh wait, maybe I see. This is a
> physical ATAPI drive on the ESX host. ESX has laundered it through as
> SCSI to the virtual BLC, stamping "ATAPI" into the vendor field.
>

This is the only configuration (using blc for the virtual CD) I could
get to work with Microlite Backup Edge on VMWare player 3.1.3.

> ...
>
> Ok, ignoring all of the above, let me just make a few comments about the
> state of OSR5-under-ESX these days.
>
> 1. There are few reasons to install a virtual SCSI adapter into an OSR5
> VM. Use IDE virtual hard disk, IDE (ATAPI) virtual CD/DVD drives.
>

I would certainly like to configure the ESXi host hardware and ESXi to
present the CD/DVD as IDE (ATAPI). The only question is how to make it
work.

On the 5.0.7 physical machine the client has switched from DVD-RAM to
DVD-R media to generate permanent archival images usable in an audit
to review data files for any day of the year (distributor of controlled
pharmaceutical products).

The desire to move to the ESXi system is to obtain better usage of the
ESXi hardware capacity and obtain performance increase due to the
use of SSD drives in the ESXi host over that available on the 5.0.7
physical hardware. What little testing we did shows report time of
37 seconds vs four minutes on the physical 5.0.7 system.

I'm not wild about switching from 5.0.7 to 5.0.7V* with the switch to the
yearly subscription model. That may change if testing shows a clear
superiority in performance moving to the ESXi environment. A better
option may be to swap out the Adaptec SCSI RAID for SATA-RAID with SSD
drivers in the physical 5.0.7 hardware.

[Bela Lubkin]

Steve M. Fabac, Jr. wrote:

> I tried both options without success. Connect at power up still fails
> the dtype /dev/cd0 test after booting the 5.0.7 VM and placing a
> valid media (DVD disk with data on it) in the ESXi host machine DVD
> drive. Clicking the CD icon in the vShpere console session for the SCO
> VM offers many options to "connect" the drive. One item that passed
> the dtype /dev/cd0 test was selecting "connect -> D:" and placing the
> disk in the DVD drive on the Windows XP system where the Vsphere client
> software is running. Clicking the CD icon and selecting "connect -> ESXi"
> host drive with the DVD media in the drive takes a long time to execute
> and then fails the dtype /dev/cd0 test.
>
> Perhaps ESXi is not configured correctly? The Host device offered is:
>
> /vmfs/devices/cdrom/mpx.vmhba36:C0:TO:L0

That's a plausible path for ESXi 4.1 or 5. So don't know why it isn't
working for you. When you try to access the drive from the VM, do you
hear anything or see any blinkenlights on the drive?

> > 1. There are few reasons to install a virtual SCSI adapter into an OSR5
> > VM. Use IDE virtual hard disk, IDE (ATAPI) virtual CD/DVD drives.
>
> I would certainly like to configure the ESXi host hardware and ESXi to
> present the CD/DVD as IDE (ATAPI). The only question is how to make it
> work.

I'll have to play with this next time I'm at my ESXi test box. I've
been away from work for several days dealing with a family member's
horrible medical situation :(

> I'm not wild about switching from 5.0.7 to 5.0.7V* with the switch to the
> yearly subscription model. That may change if testing shows a clear
> superiority in performance moving to the ESXi environment. A better
> option may be to swap out the Adaptec SCSI RAID for SATA-RAID with SSD
> drivers in the physical 5.0.7 hardware.

I argued with the SCO people shortly before 507V(1) was released, and
many times thereafter, that the subscription model was just death.
Nobody is "wild" or even "vaguely OK" with switching to the subscription
model for their supposedly "mission critical" OS which is supplied by a
bankrupt company. Or now a new company that doesn't seem to be doing
much and has built up no sense of forward continuity.

So yeah, I don't blame you. The one thing that heartens me is the
absolute certainty that if the source of those licenses goes completely
out of business, information will leak out on how to break the
licensing. (Either leak out of the tech people from the company, or be
"cracked" by people in the field. (*)) So it's like putting the noose
around your own neck, but knowing that the rope is rather frayed.

(*) At one point, the SCO people defending the subscription model
claimed that they would issue a statement to that effect. Basically a
promise "now" that if licenses were no longer commercially available,
they would release a key generator program "then". I somehow doubt such
a statement was actually issued. At least I never saw it. If someone
runs across it, please do post... Of course such a promise is
essentially worthless, but if I saw an official one + dead company (no
more license issuing authority) + no officially released keygen, I'd
feel 100% in the right cracking it and releasing the info...

>Bela<