[Bug]: Servers created with public_net disabled are delayed by 4 minutes on ubuntu 24.04 #1074
Comments
This comment has been minimized.
This comment has been minimized.
|
I'm experiencing the same thing. Is possible to disable metadata datasource using cloud-init as a workaround? I've tried few things without success. |
|
same here with not even dns is working |
|
I was able to reproduce this bug using ubuntu 24.04, but didn't had any problems with debian-12. terraform {
required_providers {
hcloud = {
source = "hetznercloud/hcloud"
version = "1.49.1"
}
}
}
data "hcloud_ssh_keys" "all" {}
resource "hcloud_network" "main" {
name = "main"
ip_range = "10.0.0.0/16"
}
resource "hcloud_network_subnet" "main" {
network_id = hcloud_network.main.id
type = "cloud"
network_zone = "eu-central"
ip_range = "10.0.1.0/24"
}
resource "hcloud_server" "bastion" {
name = "bastion"
server_type = "cx22"
image = "debian-12"
ssh_keys = data.hcloud_ssh_keys.all.ssh_keys.*.name
network {
network_id = hcloud_network.main.id
ip = "10.0.1.1"
alias_ips = []
}
}
resource "hcloud_server" "host" {
depends_on = [hcloud_network.main, hcloud_network_subnet.main]
name = "host"
server_type = "cx22"
image = "debian-12"
ssh_keys = data.hcloud_ssh_keys.all.ssh_keys.*.name
public_net {
ipv4_enabled = false
ipv6_enabled = false
}
network {
network_id = hcloud_network.main.id
ip = "10.0.1.2"
alias_ips = []
}
}On the host, while debian logs: Ubuntu logs: We can see the 2 x 2 minutes timeouts. While the server are created as fast as expected, cloud init does wait 4 additional minutes. I'll report this to the relevant team, and report back here. |
|
We believe that the source of the problem is a race between hc-utils, and cloud-init. We have an idea how to implement a fix, but it is not an easy one and will take some time to be fixed. I will try to come up with a ready to use workaround. EDIT: There is not easy workaround (directly from the user data) as far as I am aware. One may create a snapshot of the server after changing the network configuration on the disk (e.g setting https://cloudinit.readthedocs.io/en/latest/reference/network-config-format-v2.html#optional-bool or maybe disabling the |
jooola
changed the title
|
Actually I'm already buiding new Ubuntu 24.04 cax11 snapshot weekly with latest updates based on this Packer example: https://github.com/hetznercloud/packer-plugin-hcloud/tree/main/example/docker. I'm only using upgrade.sh and cleanup.sh scripts. I tried to add |
|
@perttuk In that case you may wanna try the other trick, by setting the |
|
I'm not sure if I did what I was supposed to, but after modifying Packer file to upload network:
ethernets:
enp7s0:
optional: trueI was managed to cut start time in half: Creation complete after 2m28s I forgot to add that this slow start also happens with public_net enabled when using snapshot. |
|
@perttuk Yes, you implemented what I had in mind. Does the system logs show anything relevant? |
|
I already destroyed the last server and now when re-created it it took 4m28s. Here is some logs. Hope these helps. I executed Terraform at 16:59 First journalctl log appears 2,5 mins later System is up after 10 seconds Then it idles for a while before starting ssh That's about it. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
What happened?
Servers created with public_net disabled are delayed by 4 minutes at startup because "Failed to wait for network: Unexpected error while running command."
Trying to create an instance in an air-gapped environment (no public ip) using the code below:
The startup time of this instance is 4 minutes longer than if i enable ipv4_enabled.
Here is the output of /var/log/cloud-init-output.log
What did you expect to happen?
I expect a flag or someway to prevent waiting for network if there's no ip address
Please provide a minimal working example
The text was updated successfully, but these errors were encountered: