Rerunning cloud-init

[chei33]

Hi,

I have a VM set up with cloud-init according to the instructions at https://pve.proxmox.com/wiki/Cloud-Init_Support using the Ubuntu Focal cloud image. This all worked flawlessly the first time I did it (thanks!). However, later I wanted to change the network configuration without recreating the VM. I ran

qm set 100 --ipconfig0 ...
qm set 100 --net0 ...

with the new configuration. The first to update cloud-init, and the second to update kvm. However, when the VM boots, the netplan configuration file doesn't get updated, and as far as I can tell, cloud-init isn't "fully" run. So the VM was still trying to use the old config internally. What seemed to work was to run cloud-init clean and cloud-init init, then reboot (twice?), and this caused the new cloud-init config to "take".

The documentation doesn't mention whether it's supposed to be possible to rerun cloud-init when the config changes; I guessed that it's supposed to just work, and I can't find other people running into this problem! My VM config is

root@ainosticssrv02:~# qm config 100
agent: 1
bios: ovmf
boot: c
bootdisk: scsi0
cipassword: **********
cores: 16
efidisk0: local-lvm:vm-100-disk-1,size=4M
hostpci0: 81:00.0,x-vga=on
ide2: local-lvm:vm-100-cloudinit,media=cdrom
ipconfig0: ip=192.168.140.100/24,gw=192.168.140.1
memory: 131072
name: sandbox
net0: virtio=46:07:A1:74:69:EC,bridge=vmbr1
numa: 0
onboot: 1
scsi0: local-lvm:vm-100-disk-0,size=53452M
scsihw: virtio-scsi-pci
serial0: socket
smbios1: uuid=e3b0a5a4-a1a3-44fc-8766-2e250bdc4a4a
sockets: 1
sshkeys: <redacted>
vga: qxl
vmgenid: b73354d9-cb3f-4eb5-a8ab-35bff966644b

Could the use of ovmf (needed for GPU PCI passthrough) be responsible (I changed this after originally creating the VM)?

I realised later that the gateway I was setting was wrong - could it be refusing to apply the new networking because the config was wrong? There didn't seem to be anything useful in the logs.

 

[mira]

Network config should be applied once per instance. Did you boot it once before with the config being exactly that?
The 'instance-id' is basically the hash of the user config and the network config concatenated. cloud-init checks the current instance-id against all previously known ones to see if it has to rerun the different systems and modules, not just the previous instance-id, which leads to strange behavior sometimes.

In this case the only solution currently is to run `cloud-init clean` and then reboot.

 

[chei33]

Hi mira,

Thanks - that is useful information to know. It's possible that I booted it once with this config, but with some other problem which caused it to fail to boot, but the config was still registered by cloud-init but not stored in netplan, and so never "took".

Is there a way to force it to re-apply the cloud-init config, e.g. by passing a kernel parameter?

There is this post I found, https://www.burgundywall.com/post/using-cloud-init-to-set-static-ips-in-ubuntu-20-04, which suggests that the whole process with Ubuntu 20.04 and cloud-init with Proxmox is a bit messy. He's trying to something a bit more convoluted than me, I think, involving booting the image, setting it up, and then templating it, but it might be useful for other people who run across this thread, so I reference it here anyway.