Fedora CoreOS deployment by OVA sometimes fails because ignition config is not taken

Project Discussion
#1

So my deployment is as so:
I deployed a new fedora coreOS machine on ESXi with a custom igntion config.
i then installed some stuff when finished what i did is:

sudo rm -f /etc/machine-id

and then

sudo touch /boot/ignition.firstboot

and then zeroed all the other stuff and power off the machine.
I then export this machine to its vmdk and with an ovf file i create the manifest file and then create OVA.

My problem is this:
I then try to deploy a new machine in ESXi using a new ignition config string.
for some reason - ONLY SOMETIMES - the OS fails to use the ignition config at the first boot and it says that it already ran a boot once before (referencing the old ignition config that i said above). so usually i delete this machine then deploy a new machine with the same OVA i created and this time it does succeed.

Is this a bug?
sometimes it succeeds only after multiple tries…

I’m almost hopeless…

Thanks, Ofir

#2

Hey Ofir. We don’t really support running Ignition more than once on a machine, which is technically what you’re doing here. Maybe you can explain what you’re doing when you customize the instance and then export to a VMDK to then use for other machines.