Bug 1305177

Created attachment 1121525 [details]
The error

Created attachment 1121526 [details]
the pre-error screen

Looking at the bug I assume that the system you are using is first trying to bot using EFI or fallback to BIOS otherwise.
The old rhev-hypervisor7-7.2-20160120.0 did not have EFI support, and was likely installed in BIOS mode.

The new rhev-hypervisor7-7.2-20160126.0 does support EFI, and thus the system booted the ISO in EFI mode.

I see two options:

1. Disable EFI on the system you are using
2. Keep EFI enabled and reinstall the host

Created attachment 1130867 [details]
upgrade exception after disabling uefi

Created attachment 1130869 [details]
upgrade exception screen after disabling uefi

Hi Fabian,
I apologize it took me such a long time to reply.
I disabled UEFI and tried to upgrade again.
After reboot it brought me to the upgrade page again, so I believe the upgrade failed. So do the logs. From ovirt-node.log:

~~~
2016-02-26 17:03:57,351      ERROR Installer transaction failed
Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/ovirt/node/installer/core/progress_page.py", line 126, in __run
    self.progress_plugin.dry_or(do_commit)
  File "/usr/lib/python2.7/site-packages/ovirt/node/plugins.py", line 188, in dry_or
    return func()
  File "/usr/lib/python2.7/site-packages/ovirt/node/installer/core/progress_page.py", line 120, in do_commit
    tx_element.commit()
  File "/usr/lib/python2.7/site-packages/ovirt/node/installer/core/progress_page.py", line 284, in commit
    boot_setup = install.ovirt_boot_setup()
  File "/usr/lib/python2.7/site-packages/ovirtnode/install.py", line 667, in ovirt_boot_setup
    if not self.grub2_install():
  File "/usr/lib/python2.7/site-packages/ovirtnode/install.py", line 339, in grub2_install
    _functions.remove_efi_entry(self.efi_name)
  File "/usr/lib/python2.7/site-packages/ovirtnode/ovirtfunctions.py", line 1805, in remove_efi_entry
    for entry in efi.list_entries():
  File "/usr/lib/python2.7/site-packages/ovirt/node/utils/system.py", line 899, in list_entries
    lines = self._efibootmgr([("verbose", None)])
  File "/usr/lib/python2.7/site-packages/ovirt/node/utils/system.py", line 870, in _efibootmgr
    return self._call(cmd)
  File "/usr/lib/python2.7/site-packages/ovirt/node/utils/system.py", line 873, in _call
    return process.check_output(cmd)
  File "/usr/lib/python2.7/site-packages/ovirt/node/utils/process.py", line 159, in check_output
    stdout = subprocess.check_output(*args, **kwargs)
  File "/usr/lib64/python2.7/subprocess.py", line 575, in check_output
    raise CalledProcessError(retcode, cmd, output=output)
CalledProcessError: Command '['efibootmgr', '--verbose']' returned non-zero exit status 2
~~~

Fabian, let me know what should I do next.
I'll keep the host as is for now.

I believe I can reinstall, but question is if this problem would not occur upgrading 3.5 hosts to 3.6 because of this problem.

I just tried upgrading with beta3 and same problem is there. Upgrade fails with same error.

Marina, could you please try reinstalling the host with 3.5 and then upgrade to 3.6 beta3 to see if the bug exists?

Can the following scenario be tested:

1. Install RHEV-H 3.5 on a host with EFI and CSM enabled (fallback to BIOS if EFI is not found)
2. Media upgrade to RHEV-H 3.6 rc1 or GA

we tested the following for upgrade, all passed:

1. UEFI machine upgrade
  1.1 Upgrade from rhev-hypervisor7-7.1-20150420.0.iso to rhevh-7.2-20160301.1.el7ev.iso on UEFI machine - PASS
  1.2 Upgrade from rhev-hypervisor7-7.1-20150603.0.iso to rhevh-7.2-20160126.0.el7ev.iso on UEFI machine(hp-bl460cg9) - PASS

2. BIOS machine upgrade
  2.1 Upgrade from 7.2_3.6 Beta3 rhevh-7.2-20160126.0.el7ev.iso and RC1(rhevh-7.2-20160225.0.el7ev.iso) to rhevh-7.2-20160301.1.el7ev.iso - PASS
  2.2 Upgrade from 7.2_3.5.8 to rhevh-7.2-20160301.1.el7ev.iso - PASS
  2.3 Upgrade from 7.1_3.5.z latest rhevh-7.1-20151015.0 to rhevh-7.2-20160301.1.el7ev.iso - PASS

Thanks for the clarification.

But how is the EFi machine configured?
Is EFi only configured or EFI with CSM fallback?

UEFI machine upgrade on hp-bl460cg9 server:

1. Both enable UEFI and CSM, all the original testing base on this scenario: 
   1.1 Upgrade from rhev-hypervisor7-7.1-20150420.0.iso to rhevh-7.2-20160301.1.el7ev.iso on UEFI machine - PASS
   1.2 Upgrade from rhev-hypervisor7-7.1-20150603.0.iso to rhevh-7.2-20160126.0.el7ev.iso on UEFI machine(hp-bl460cg9) - PASS

2. Only enbale UEFI, CSM disable, I tested this scenario just now:
  Upgrade from 7.2-20160126.0.el7ev.iso to rhevh-7.2-20160301.1.el7ev.iso on UEFI machine(hp-bl460cg9) - PASS

(In reply to Fabian Deutsch from comment #11)
> Can the following scenario be tested:
> 
> 1. Install RHEV-H 3.5 on a host with EFI and CSM enabled (fallback to BIOS
> if EFI is not found)
> 2. Media upgrade to RHEV-H 3.6 rc1 or GA

I enabled both UEFI and CSM, installed 3.5.8 and then tried upgrading - same error as the one that I opened the bug for.

<rbarry_> You need to wipe it (reinstall/uninstall from an ISO)
 If it's booted in EFI the 2nd time, it's going to try to mount the EFI system partition, which won't exist
 So the upgrade needs to be legacy boot as well
 We'll have to tell customers to reinstall legacy systems upgraded to EFI (or just upgrade from a legacy boot), but I think that already got documented when EFI was added the first time
<mku> rbarry_, so, I tried with legacy only and it failed for another reason
 rbarry_, it is ok with me
 rbarry_, do you have any idea where it was documented maybe?
<rbarry_> Probably a kbase article. I'll see if I can find one
<mku> fabiand_afk, following this conversation, are you ok if I close the bug with a statement that if UEFI is in use, reinstall is needed?


I'll close the bug then.

https://access.redhat.com/solutions/2190591