Test: Install Server KVM x86_64 image (20261009) #216

Open
opened 2026-07-15 16:05:35 +00:00 by pboy · 42 comments
Owner

Install Server Virtualization Support and a Server KVM image following
Creating a virtual machine using Fedora Server Edition disk image

Links in issue 203

Install Server Virtualization Support and a Server KVM image following [Creating a virtual machine using Fedora Server Edition disk image ](https://docs.fedoraproject.org/en-US/fedora-server/virtualization/vm-install-diskimg-fedoraserver/) Links in [issue 203](https://forge.fedoraproject.org/server/tickets/issues/203)
korora self-assigned this 2026-07-16 14:58:11 +00:00
pboy changed title from Test: Install Server KVM x86_64 image (20260715) to Test: Install Server KVM x86_64 image (20260721) 2026-07-24 06:13:44 +00:00
pboy changed title from Test: Install Server KVM x86_64 image (20260721) to Test: Install Server KVM x86_64 image (20260724) 2026-07-24 07:14:06 +00:00
pboy changed title from Test: Install Server KVM x86_64 image (20260724) to Test: Install Server KVM x86_64 image (20260731) 2026-08-03 20:46:17 +00:00
pboy self-assigned this 2026-08-05 15:47:43 +00:00
Member

As of the 20260817 image this is working

As of the 20260817 image this is working
korora changed title from Test: Install Server KVM x86_64 image (20260731) to Test: Install Server KVM x86_64 image (20260817) 2026-08-20 02:48:10 +00:00
Author
Owner

Question: Is the wrong type of partition 4 fixed? At least until now it was Linux, but has to be LVM

Question: Is the wrong type of partition 4 fixed? At least until now it was Linux, but has to be LVM
Member

Partition 4 is listed as "Linux root (x86_64)"

but it's mounted on /dev/mapper/systemVG-LVroot

So it appears that it's an LVM, but the partition isn't correct.

Partition 4 is listed as "Linux root (x86_64)" but it's mounted on /dev/mapper/systemVG-LVroot So it appears that it's an LVM, but the partition isn't correct.
Author
Owner

OK, to the issue isn't fixed yet. As Neil (ngompa) mentioned in another thread he has contacted the Kiwi people.

It boots that way, but you get into issues with some LVM tools and probably at other occasions.

OK, to the issue isn't fixed yet. As Neil (ngompa) mentioned in another thread he has contacted the Kiwi people. It boots that way, but you get into issues with some LVM tools and probably at other occasions.
korora changed title from Test: Install Server KVM x86_64 image (20260817) to Test: Install Server KVM x86_64 image (20260820) 2026-08-20 14:47:07 +00:00
Member

The 20260820 image boots with no issues, still showing root partition as Linux root (x86_64)

The 20260820 image boots with no issues, still showing root partition as Linux root (x86_64)
goroboro changed title from Test: Install Server KVM x86_64 image (20260820) to Test: Install Server KVM x86_64 image (20260823) 2026-08-24 23:00:03 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260823.n.0_Installation
goroboro changed title from Test: Install Server KVM x86_64 image (20260823) to Test: Install Server KVM x86_64 image (20260826) 2026-08-26 15:44:38 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260826.n.0_Installation
Member

As of the 20260826 build, the image no longer boots. It hangs at "Job initial-setup.service/start"

This is possibly related to BZ https://bugzilla.redhat.com/show_bug.cgi?id=2358688#c32.

I will be trying to do some testing of this later today.

As of the 20260826 build, the image no longer boots. It hangs at "Job initial-setup.service/start" This is possibly related to BZ https://bugzilla.redhat.com/show_bug.cgi?id=2358688#c32. I will be trying to do some testing of this later today.
goroboro changed title from Test: Install Server KVM x86_64 image (20260826) to Test: Install Server KVM x86_64 image (45 Beta 1.2) 2026-09-04 10:50:39 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Beta_1.2_Installation
goroboro changed title from Test: Install Server KVM x86_64 image (45 Beta 1.2) to Test: Install Server KVM x86_64 image (45 Beta 1.3) 2026-09-09 18:00:41 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Beta_1.3_Installation
Author
Owner

Test of Beta 1.3

Installation went with two error messages, but system continues booting

  • FAILED] Failed to start systemd-repart.service - Repartition Root Disk.
    See 'systemctl status systemd-repart.service' for details.
  • [FAILED] Failed to start NetworkManager-wait-online.service - Network Manager Wait Online.
    See 'systemctl status NetworkManager-wait-online.service' for details.

At the end the system continuously shows

  • Job initial-setup.service/start running (xxmin yys / no limit)

And leaves an unusable system.

It's a regression, branched 20260912 worked.

Test of Beta 1.3 Installation went with two error messages, but system continues booting - FAILED] Failed to start systemd-repart.service - Repartition Root Disk. See 'systemctl status systemd-repart.service' for details. - [FAILED] Failed to start NetworkManager-wait-online.service - Network Manager Wait Online. See 'systemctl status NetworkManager-wait-online.service' for details. At the end the system continuously shows - Job initial-setup.service/start running (xxmin yys / no limit) And leaves an unusable system. It's a regression, [branched 20260912](https://kojipkgs.fedoraproject.org/compose/branched/Fedora-45-20260912.n.0/compose/Server/x86_64/images/Fedora-Server-Guest-Generic-45-20260912.n.0.x86_64.qcow2) worked.
Author
Owner

Regarding documentation:

All the screenshots about Cockpit VM import must receive an update. The text in this section is (mostly) OK. A review can’t do any harm, though.

I did an update of the other parts, will upload soon.

Regarding documentation: All the screenshots about Cockpit VM import must receive an update. The text in this section is (mostly) OK. A review can’t do any harm, though. I did an update of the other parts, will upload soon.
goroboro changed title from Test: Install Server KVM x86_64 image (45 Beta 1.3) to Test: Install Server KVM x86_64 image (20260912) 2026-09-13 23:00:02 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260912.n.0_Installation
goroboro changed title from Test: Install Server KVM x86_64 image (20260912) to Test: Install Server KVM x86_64 image (20260915) 2026-09-15 23:00:05 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260915.n.0_Installation
Member

I can't reproduce this as reported. I downloaded https://dl.fedoraproject.org/pub/alt/stage/45_Beta-1.3/Server/x86_64/images/Fedora-Server-Guest-Generic-45_Beta-1.3.x86_64.qcow2 , attached it to a BIOS VM, booted the VM, and it booted straight to initial-setup (console version). With initial-setup I could create an admin user, then exit initial-setup, and I was at a working system.

I didn't do the whole virt-install / cockpit thing, I just attached the image directly to a virt-manager VM and booted it, but I don't see how that would make a difference.

I can't reproduce this as reported. I downloaded https://dl.fedoraproject.org/pub/alt/stage/45_Beta-1.3/Server/x86_64/images/Fedora-Server-Guest-Generic-45_Beta-1.3.x86_64.qcow2 , attached it to a BIOS VM, booted the VM, and it booted straight to initial-setup (console version). With initial-setup I could create an admin user, then exit initial-setup, and I was at a working system. I didn't do the whole virt-install / cockpit thing, I just attached the image directly to a virt-manager VM and booted it, but I don't see how that would make a difference.
Member

I downloaded https://dl.fedoraproject.org/pub/alt/stage/45_Beta-1.3/Server/x86_64/images/Fedora-Server-Guest-Generic-45_Beta-1.3.x86_64.qcow2 and followed the instructions in https://docs.fedoraproject.org/en-US/fedora-server/virtualization/vm-install-diskimg-fedoraserver/ to create a VM from within Cockpit. It booted into initial-setup without a problem. I could create an admin user and exit initial-setup and access the working system.

I downloaded https://dl.fedoraproject.org/pub/alt/stage/45_Beta-1.3/Server/x86_64/images/Fedora-Server-Guest-Generic-45_Beta-1.3.x86_64.qcow2 and followed the instructions in https://docs.fedoraproject.org/en-US/fedora-server/virtualization/vm-install-diskimg-fedoraserver/ to create a VM from within Cockpit. It booted into initial-setup without a problem. I could create an admin user and exit initial-setup and access the working system.
Member

I downloaded https://kojipkgs.fedoraproject.org/compose/branched/Fedora-45-20260918.n.0/compose/Server/x86_64/images/Fedora-Server-Guest-Generic-45-20260918.n.0.x86_64.qcow2 on both an F44 and and F45 machine and I just stop at "Job initial-setup.service/start running" I've got one that's been sitting for around 30 minutes at this point (on the F44 install) and one at 10 minutes (the F45 install). The only thing I did to either image was use qemu-img to resize them to 20G.

I downloaded https://kojipkgs.fedoraproject.org/compose/branched/Fedora-45-20260918.n.0/compose/Server/x86_64/images/Fedora-Server-Guest-Generic-45-20260918.n.0.x86_64.qcow2 on both an F44 and and F45 machine and I just stop at "Job initial-setup.service/start running" I've got one that's been sitting for around 30 minutes at this point (on the F44 install) and one at 10 minutes (the F45 install). The only thing I did to either image was use qemu-img to resize them to 20G.
Member

Huh. I'll try with a nightly then. Possibly the fix we landed post-beta broke this case.

Huh. I'll try with a nightly then. Possibly the fix we landed post-beta broke this case.
Member

@goroboro shut down the vm and reconneed with a graphical console and inital setup appeared. It does not appear that the inital-setup service is displaying to the serial console.

still an issue, but something that can be worked around...

@goroboro shut down the vm and reconneed with a graphical console and inital setup appeared. It does not appear that the inital-setup service is displaying to the serial console. still an issue, but something that can be worked around...
Member

Oh, are you guys testing with serial console? I always test with graphical display. I'll try serial console too next time.

Oh, are you guys testing with serial console? I always test with graphical display. I'll try serial console too next time.
Member

In the past the qcow image would display inital setup to the serial console. depeending on why you are setting up a VM It may not need full fledged graphics. I still need to pull Beta 1.3 and test that and see what happens.

In the past the qcow image would display inital setup to the serial console. depeending on why you are setting up a VM It may not need full fledged graphics. I still need to pull Beta 1.3 and test that and see what happens.
Member

yes, it certainly should display on serial console, if it doesn't that's a bug.

yes, it certainly should display on serial console, if it doesn't that's a bug.
Member

Interesting... I would have thought that serial console is reserved for system log output... so using serial console to display a tui would be an interesting conflict of purpose. But I don't know history of this. I always used graphical console as well... I just happened to accidentally start a VM and open serial console and then noticed what everyone was talking about. I switched to Graphical and everything was fine.

Interesting... I would have thought that serial console is reserved for system log output... so using serial console to display a tui would be an interesting conflict of purpose. But I don't know history of this. I always used graphical console as well... I just happened to accidentally start a VM and open serial console and then noticed what everyone was talking about. I switched to Graphical and everything was fine.
Member

So I thought I could reproduce this and filed https://bugzilla.redhat.com/show_bug.cgi?id=2537105 ... but now it won't reproduce for me any more. Very confusing.

So I thought I could reproduce this and filed https://bugzilla.redhat.com/show_bug.cgi?id=2537105 ... but now it won't reproduce for me any more. Very confusing.
goroboro changed title from Test: Install Server KVM x86_64 image (20260915) to Test: Install Server KVM x86_64 image (20260918) 2026-09-18 23:00:04 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260918.n.0_Installation
Member

@korora wrote in #216 (comment):

Partition 4 is listed as "Linux root (x86_64)"

but it's mounted on /dev/mapper/systemVG-LVroot

So it appears that it's an LVM, but the partition isn't correct.

This still applies as of the 20260918 nightly build.

@korora wrote in https://forge.fedoraproject.org/server/tickets/issues/216#issuecomment-1343585: > Partition 4 is listed as "Linux root (x86_64)" > > but it's mounted on /dev/mapper/systemVG-LVroot > > So it appears that it's an LVM, but the partition isn't correct. This still applies as of the 20260918 nightly build.
Author
Owner

Well, I always tested with serial console, that's the default in Cockpit, if I see it correctly. And we need it working with plain CLI as well, that is with virt-install.

Well, I always tested with serial console, that's the default in Cockpit, if I see it correctly. And we need it working with plain CLI as well, that is with virt-install.
Member

I've been trying to reproduce with serial console and I think I did just one time, but every other try - with whatever image I tested - showed it on the serial console just fine. I guess I didn't try a VM without a graphics card at all yet, I can try that next...

I've been trying to reproduce with serial console and I think I did just *one* time, but every other try - with whatever image I tested - showed it on the serial console just fine. I guess I didn't try a VM without a graphics card at all yet, I can try that next...
pboy changed title from Test: Install Server KVM x86_64 image (20260918) to Test: Install Server KVM x86_64 image (20260921) 2026-09-21 14:33:02 +00:00
goroboro changed title from Test: Install Server KVM x86_64 image (20260921) to Test: Install Server KVM x86_64 image (20260924) 2026-09-24 23:00:03 +00:00
Member
Title updated for new build. See https://fedoraproject.org/wiki/Test_Results:Fedora_45_Branched_20260924.n.0_Installation
Author
Owner

Tested Branched 2026 09 24: Still failing

If I use CLI to install as well as Cockpit VM import tool the installation ends up in an infinite loop, following our installation guide 'Creating a virtual machine using Fedora Server Edition disk image'. The result was consistent in 3 consecutive installations.

It stops if I switch Cockpit to graphical mode. Cockpit then adds VNC to the VM configuration and restarts the VM. That configuration change is illegitimate. Fedora Server is installed headless without any graphical functionality. So, this is not a solution at all.

Also, the issue with the incorrect partition type and the missing LVM device file in /etc/lvm/devices does still exist.

Tested Branched 2026 09 24: Still failing If I use CLI to install as well as Cockpit VM import tool the installation ends up in an infinite loop, following our installation guide '[Creating a virtual machine using Fedora Server Edition disk image](https://docs.fedoraproject.org/en-US/fedora-server/virtualization/vm-install-diskimg-fedoraserver/)'. The result was consistent in 3 consecutive installations. It stops if I switch Cockpit to graphical mode. Cockpit then adds VNC to the VM configuration and restarts the VM. That configuration change is illegitimate. Fedora Server is installed headless without any graphical functionality. So, this is not a solution at all. Also, the issue with the incorrect partition type and the missing LVM device file in /etc/lvm/devices does still exist.
pboy changed title from Test: Install Server KVM x86_64 image (20260924) to Test: Install Server KVM x86_64 image (20260927) 2026-09-27 10:42:06 +00:00
Author
Owner

Just tested Branched 2026 09 27: Still failing, no progress in any of the issues at all.

Just tested Branched 2026 09 27: Still failing, no progress in any of the issues at all.
Author
Owner

@adamwill We can reproduce the issue permanently, if we use virt-install on CLI, which is the most appropriate method for Server. And we get a reproducible result if we use Cockpit in its default behavior, i.e. using the terminal in serial mode. Under some conditions you get quite an ugly menu.

Anyway, do you have an idea how to proceed? I think we can't release a non-installable distributable.

@adamwill We can reproduce the issue permanently, if we use virt-install on CLI, which is the most appropriate method for Server. And we get a reproducible result if we use Cockpit in its default behavior, i.e. using the terminal in serial mode. Under some conditions you get quite an ugly menu. Anyway, do you have an idea how to proceed? I think we can't release a non-installable distributable.
Member

Replacing /usr/lib/systemd/system/initial-setup.service with the version that was in F44 would resolve this issue. I don't know how to make this happen in the qcow2 build. But I edited a local qcow and tested and this works.
The key difference is that in the f45 package the initial-setup is run inside kmscon on -v 7.
The f44 version just uses whatever tty is available.

Replacing `/usr/lib/systemd/system/initial-setup.service` with the version that was in F44 would resolve this issue. I don't know how to make this happen in the qcow2 build. But I edited a local qcow and tested and this works. The key difference is that in the f45 package the initial-setup is run inside kmscon on -v 7. The f44 version just uses whatever tty is available.
Member

After further exploration. This seems to be an issue with logging output clobbering the console so that people are finding it difficult to interact with the menu. We could keep the 45 initial-setup systemd file, but we could add the following drop-in:

/etc/systemd/system/initial-setup.service.d/quiet-console.conf

[Service]
ExecStartPre=/bin/kill -SIGRTMIN+21 1
ExecStartPost=/bin/kill -SIGRTMIN+20 1

This basically just stops the log output to the serial console while initial-setup is running, and then starts it again after it is finished running. The lines come from the f44 service file, and seem to work well in test.

After further exploration. This seems to be an issue with logging output clobbering the console so that people are finding it difficult to interact with the menu. We could keep the 45 initial-setup systemd file, but we could add the following drop-in: /etc/systemd/system/initial-setup.service.d/quiet-console.conf [Service] ExecStartPre=/bin/kill -SIGRTMIN+21 1 ExecStartPost=/bin/kill -SIGRTMIN+20 1 This basically just stops the log output to the serial console while initial-setup is running, and then starts it again after it is finished running. The lines come from the f44 service file, and seem to work well in test.
Member

Well, I think we could consider just putting those lines back universally, then. @jfalempe removed them in github.com/rhinstaller/initial-setup@01b6b9e43a , presumably on the theory that, because the TUI is now running on a dedicated VT, they weren't needed any more; but that ignores the serial terminal case. WDYT @jfalempe ?

Well, I think we could consider just putting those lines back universally, then. @jfalempe removed them in https://github.com/rhinstaller/initial-setup/commit/01b6b9e43a044710bc40331e176d4d619316f65e , presumably on the theory that, because the TUI is now running on a dedicated VT, they weren't needed any more; but that ignores the serial terminal case. WDYT @jfalempe ?

I removed this two lines because it has no effect when using kmscon.
As you probably override this service to run on a serial line, it might be better to keep those commands in the overlay.
It doesn't make sense to stop the console logging when the service starts kmscon.

I removed this two lines because it has no effect when using kmscon. As you probably override this service to run on a serial line, it might be better to keep those commands in the overlay. It doesn't make sense to stop the console logging when the service starts kmscon.
Member

@jfalempe anyone might be using a serial console on any Fedora system. We should always consider the serial console use case a valid one that should work. I think the lines should be restored. I don't think anyone is overriding anything to run on the serial console, it's always done that, I assume something in initial-setup itself knows to write to and read from serial consoles?

@jfalempe anyone might be using a serial console on any Fedora system. We should always consider the serial console use case a valid one that should work. I think the lines should be restored. I don't think anyone is overriding anything to run on the serial console, it's always done that, I assume something in initial-setup itself knows to write to and read from serial consoles?

ok, I think I understand, is that when kmscon is starting run-initial-setup script, this script can also listen to the serial port.

I though that you needed to override the initial-setup ExecStart to start something on the serial line explicitly.
So in this case I can restore the two lines.
Sorry for breaking this use case.

ok, I think I understand, is that when kmscon is starting `run-initial-setup` script, this script can also listen to the serial port. I though that you needed to override the initial-setup ExecStart to start something on the serial line explicitly. So in this case I can restore the two lines. Sorry for breaking this use case.
Member

Yes! Thank you, that would probably resolve the issue for users connecting directly to the serial port.
And no problem, it wasn't immediately apparent as to what was going on. But I think this would totally resolve this issue.

Yes! Thank you, that would probably resolve the issue for users connecting directly to the serial port. And no problem, it wasn't immediately apparent as to what was going on. But I think this would totally resolve this issue.
I have submitted https://github.com/rhinstaller/initial-setup/pull/167
pboy changed title from Test: Install Server KVM x86_64 image (20260927) to Test: Install Server KVM x86_64 image (20261003) 2026-10-03 17:45:08 +00:00
Author
Owner

I just installed the 20261003 KVM.

The initial-setup menu is fine now.

LVM partition type is still wrong and LVM is still not correctly set up (the /etc/lvm/devices/system.devices file is still missing).

The virtual disk image size has been increased to 40GB which is much more a reasonable size.

Would be nice to get the rootVG/file system get enlarged to our 15 GB default value, too. Would make ISO and KVM perfectly in sync.

I just installed the 20261003 KVM. The initial-setup menu is fine now. LVM partition type is still wrong and LVM is still not correctly set up (the /etc/lvm/devices/system.devices file is still missing). The virtual disk image size has been increased to 40GB which is much more a reasonable size. Would be nice to get the rootVG/file system get enlarged to our 15 GB default value, too. Would make ISO and KVM perfectly in sync.
pboy changed title from Test: Install Server KVM x86_64 image (20261003) to Test: Install Server KVM x86_64 image (20261009) 2026-10-09 18:40:23 +00:00
Author
Owner

I tested the 20261009 KVM.

  • the initial-setup issue is back again, same behavior as pre 20261003
  • the max size of the virtual disk is back to 10 GB
  • the type of the LVM partition (partition 4) is still wrong (Linux Root instead of LVM2)
  • the LVM devices table is still missing (empty auto-import-rootvg tag file instead of system.devices file)

Basically it is a regression.

I tested the 20261009 KVM. - the initial-setup issue is back again, same behavior as pre 20261003 - the max size of the virtual disk is back to 10 GB - the type of the LVM partition (partition 4) is still wrong (Linux Root instead of LVM2) - the LVM devices table is still missing (empty auto-import-rootvg tag file instead of system.devices file) Basically it is a regression.
Sign in to join this conversation.
No milestone
No assignees
5 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
server/tickets#216
No description provided.