This message was deleted.

Hey, it's me! 😆 So, this is a bit weird. I don't think it is related to that backport, anyway I cannot explain how that could happen 🤔 I tried it on a local setup with Elemental 1.5.3 with no luck in reproducing it. So, the "name" is the name of the

MachineInventory

associated to the host. It is picked up at the first registration and is set to the

MachineRegistration

machineName

field (or to a "m-$UUID" default name if not specified). In the MachineRegistration below, the

machineName

is set to "${System Data/Runtime/Hostname}", i.e., it will pick the hostname of the host at registration time. So, at the very first registration of the host, a new

MachineInventory

is created with a name equal to the

machineName

. The

MachineInventory

name is set as the static hostname at k3s/RKE2 provisioning time. Now, the label

hostname

you added, which is again equal to "${System Data/Runtime/Hostname}", i.e., it takes the current hostname of the host, is updated and re-evaluated every time the registration happens. Having a "m" there means that the host had an empty hostname at the last registration time. Could it be an unfortunate clash in the emulated TPM seed from another host? 🤔 And why did the hostnamctl command didn't worked? Could it be that the vmtoolsd returned nothing? 🤔 I cannot really see how it could have ended up like that. It may be worth a bug if you saw it more than once, please share also the elemental version (besides sharing the MachineRegistration and the SeedImage as you already did, quite important).

ok, the cluster is already deleted and i moved to 2.0.4. jep, maybe it was tpm related.

well, mine is just a wild guess, there could be a bug/race somewhere... so thanks for reporting this kind of issues, if it happens again please share!

I delete all vms of a working cluster, removed them from the inventory too, then I recreated the vms with terraform: redacted domainsuffix which somehow is added after the node is adopted. i think this is new in 2.0.4?