D370 fails to boot from alternate disk without interaction

[David Burgess]

I am practicing my recovery skills and have damaged the primary side of the mirror (c0t5d0) by dd'ing over the header. The other side of the mirror is c0t8d0. The alternate boot path is set to this.

The AUTO file on both disks contains "hpux -lq" so even though c0t5d0 is broken, c0t8d0 should boot. On boot the D370 fails to boot c0t5d0 (as expected). It then fails to search and find the alternate boot disk.

This is what I have :-

# dd if=/dev/vg00/lvol3 of=/dev/rdsk/c0t5d0 bs=1024 (Where c0t5d0 is the primary root disk)


Reboot the server.

# shutdown -ry 0

On reboot

IPL error; bad LIF magic.
.... FAILED.

Trying manufacturing default devices
------------------------------------

Cannot find ENTRY_TEST.

Failed to initialize.

ENTRY_INIT
Status = -7

00000080 00000000 00000000 00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
0000350B 00000012 E80818B2 02002080 804000DC 00000000 0021F000 00000000
9E030000 0010295C 00000000 FFFFFFFF 00000000 00000000 00000000 00000000

Failed to initialze.

.... FAILED.

Terminating selection process.
No boot device found.
.
.
.
.
.

-------

Main Menu: Enter Command > sea

Path Number Device Path (dec) Device Type
----------- ----------------- -----------
P0 8/4.11 Random access media
P1 8/4.10 Random access media
P2 8/4.9 Random access media
P3 8/4.8 Random access media
P4 8/4.5 Random access media
P5 8/16/5.2 Random access media
P6 8/16/5.0 Sequential access media
P7 8/16/6.0 LAN Module

Main Menu: Enter command > bo alt
Interact with IPL (Y, N, or Cancel)?> n

Booting...
Boot IO Dependent Code (IODC) revision 152

HARD Booted.

ISL Revision A.00.43 Apr 12, 2000

.
.
.
ISL booting hpux -lq


Boot
: disc (8/4.8.0;0)/stand/vmunix

Then boots ok.
Any ideas on how I can make this D370 find the alternate disk at 8/4.8 and boot without any interaction?

Regards,

Dave.

[Hoefnix]

When creating the mirror bootdisk you can set the -lq:
mkboot -a 'hpux -lq' /dev/dsk/[disknumber]

When the alt-boot-path is set correct it should boot without interaction.

Regards,

Peter

[Chris Wilshaw]

Try running the setboot command;

>setboot
Primary bootpath : 0/0/1/1.2.0
Alternate bootpath : 0/0/2/0.2.0

Autoboot is ON (enabled)
Autosearch is ON (enabled)

You can then use the flags

-p to set primary path
-a to set alternate path
-b autoboot on/off
-s autoserch on/off

[David Burgess]

Hi,

setboot is correct. Autoboto and Autosearch are on.

I used mkboot -a "hpux -lq" /dev/rdsk/c0t5d0 and mkboot -a "hpux -lq" /dev/rdsk/c0t8d0.

So all looks like it should work. Strange!

Regards,

Dave.

[Kent Ostby]

Dave --

The Alternate Boot Disk / AUTOSEARCH features will only kick in if the system fails to detect a bootable disk drive for the Primary.

I'm not sure if you were trying to simulate a "broken disk" with the dd , but the system at this point still believes that the primary disk is a "good disk" at the time it is deciding between primary/alternate/autosearch.

Subsequent to it making the decision, it boots up, but there is no way for the HW to know that the LIF file is blown at the time it is making its decision.

In other words, its working the way it should work.

The primary/alternate/autosearch sequence works only when the HW can't see the disk in question. So in this case, if you uncabled the 0/1 bus, I would expect your system to boot the alternate and at least attempt to come up.

Best regards,

Kent M. Ostby

[Bill Hassell]

The processor ROMs have no to determine the validity of a boot disk. The code looks for a working primary disk and if true, reads the boot portion into RAM. If it does not work as expected, you'll see the error message you got. As Kent mentioned, a mirrored system is for hardware failures, not trashed information. Popping out the primary disk from your D370 is the best way to simulate a failure. Since these are hot-replaceable disks, you can remove one pr the other while the system is running. You'll see the error logged in /var/adm/syslog/syslog.log and the system will continue (there will be a short pause while the driver waits for the missing disk),

[David Burgess]

Thanks for your help. I've tried pulling out c0t5d0 (8/4.5) and rebooting.

The system sees it's missing, fails to boot and then searches for the manufacturing default devices. It then fails to find one and the boot fails. I end up at the Main Menu prompt.

pa shows alt at 8/4.5

bo alt suceeds.

Any ideas?

Regards,

Dave.

[Kent Ostby]

David -- can you post the output of the "setboot" command.

Thanks,

Kent Ostby

[David Burgess]

# setboot
Primary bootpath : 8/4.5.0
Alternate bootpath : 8/4.8.0

Autoboot is ON (enabled)
Autosearch is ON (enabled)

[Kent Ostby]

When you pull the disk at 8/4.5 and it fails to boot, does it show you 8/4.8 in a list of devices ?

I understand that you can type "boot alt" and it will boot, but does it show 8/4.5 when it looks for the devices or not ?

Might be a case, again where pulling out 8/4.5 is causing the whole bus to be unusable.

[Patrick Wallek]

Dave,

I've got a setup that is virtually identical to yours so I just did some tests. The machine I tested on is also a D370 with 2 disks, 1 in 8/4.5.0 and 1 in 8/4.8.0. They are mirrored boot disks.

I initially had my machine, set so that the Primary boot disk was 8/4.8.0 and the Alternate was 8/4.5.0. I shut the machine down, pulled out the Primary disk (8/4.8.0) and rebooted. The machine CAME UP FINE!!! It detected that the primary disk was not there, searched for an alternate and found 8/4.5.0 and booted fine!

So the next thing I did was I switched my boot paths so that the primary is now 8/4.5.0 and the alternate is 8/4.8.0. I then shut the machine down, pulled out the primary boot drive (8/4.5.0) and rebooted. The machine DID NOT boot. Unfortunately I did not do an 'sea' at the BCH prompt to see what it would find, but I seriously doubt that it would see the disk at 8/4.8.0.

So my conclusion is that there is something in the firmware or BCH of the D box that prevents it from behaving as we would expect. If the disk at 8/4.5.0 was dead and not just pulled out, it might behave differently, I don't know.

I think based on this experiment that I may have to look at which drive is the primary boot drive on ALL my D series boxes.

Try the test for yourself and see what happens.

# setboot -p 8/4.8.0
# setboot -a 8/4.5.0
# setboot
Just to verify

# shutdown -ry 0

Watch the LCD display on the box and when it starts doing its self TEST, pull out the disk at 8/4.8.0 and sit back and watch.

[David Burgess]

Kent,

Do you mean the manufacturing default devices? If so it doesn't list anything.

Only "FAILED" and "No boot device found"

After at the Main Menu: prompt, sea picks up all remaining devices including 8/4.8

Regards,

Dave.

[David Burgess]

Patrick,

I swapped over the boot devices using the pa command at the Main Menu prompt. I rebooted. It started from 8/4.8. I then pulled 8/4.8 and rebooted. The same thing happened again. It tried primary and failed. Went for the manufacturing default devices and FAILED.

Again sea finds the remaining disks.
bo alt boots 8/4.5 and we're off.

So it seems that whichever disk is pri or alt it still fails to boot with the primary disk pulled.

Now I'm worried!

Regards,

Dave.

[Patrick Wallek]

Dave,

That is very interesting. I wonder why it booted normally on my D370 when my primary was 8/4.8.0 and not 8/4.5.0 and it is not behaving well at all on yours.

Perhaps a firmware issue? I am at PDC 38.40.

[David Burgess]

Patrick,

It's weird. It is also at PDC 38.40.

I've tried resetting to default values.
I have toggled autosearch, autoboot, fastboot and it makes no odds.

I've swapped pri and alt and pulled the corresponding primary disk each time.

When it looks for the manufacturing device it flashes the cdrom light.

The cdrom is 8/16/5.2

Now I'm confused!!!

Dave.

[David Burgess]

Still looking.....

This is interesting. I've put the same 2 disks in another D370 server. This time a 2 way. Both are at PDC 38.40. The 2 way will find the alternate boot disk if the primary is set to 8/4.8 and the alt set to 8/4.5.

If I pull 8/4.8 it will look at the manufacturing default devices and boot from 8/4.5. The 1 way won't.

Also I know that if I have a hass with the alternate boot disk at 10/12.5.0 and the primary at 8/4.5 it also fails to boot alt and I get woken up early in the morning! That's what started me on this :-(

Is there a way to view or modify the manufacturing default devices?

Regards,

Dave.

[Patrick Wallek]

Now this is very interesting indeed Dave. The D370 I did my tests on yesterday is a 2-way machine as well.

I am also getting very curious about this.

Anyone from HP care to weigh in on this issue?

[David Burgess]

I've opened a call with HP. Back with an update when I have one.

Regards,

Dave.

[David Burgess]

I have an answer from HP. It appears that :-

"The 'ALT' is not a backup mechanism for PRI. It is only a definition for a second bootable device" (the D, K, and T can't autoboot from any alternate path except the default path (8/4/5) when the primary path is unavailable)

This is a limitation of the PDC code and although there have been Enhancement
Requests submitted ("Alternate Boot if Primary fails") this does not appear to
have been rolled out in any PDC-update beyond 38.40 which is the latest for a D class server.

There is a work around which is to make the alternate boot path 8/4/5. Then if the primary path fails it will find the alternate path as it is default. From testing it appears that you need to have a 2 cpu server to stand a chance!

Looks like I'll be getting out of bed if this happens again!!

Thanks for all your help.

Regards,

Dave.

[Nancy_32]

I have a problem with a SCSI disk that would not clone in a cloning system we use. I decided to use the dd command to copy the good disk to a new hard drive.

Should I have to mount the target disk? Do I have to format the disk? The disk is bootable because I was able to boot up the disk.

Thanks,

Nancy L.