[2.8.0] - load_balancer name is already used #1028

K-MeLeOn · 2023-10-14T12:29:56Z

K-MeLeOn
Oct 14, 2023

Hi everyone !

After upgrade to the last version (2.8.0), I update my kube.tf file than just some commentary was added, I use terraform apply --auto-approve command and got this error :

module.kube-hetzner.hcloud_load_balancer.cluster[0]: Creating...
╷
│ Error: name is already used (uniqueness_error)
│
│ with module.kube-hetzner.hcloud_load_balancer.cluster[0],
│ on .terraform/modules/kube-hetzner/init.tf line 1, in resource "hcloud_load_balancer" "cluster":
│ 1: resource "hcloud_load_balancer" "cluster" {
│
╵

No config change in the kube.tf file, I don't know where to search, any idea ?

Thanks for you time !!

Answered by mysticaltech

Oct 17, 2023

@K-MeLeOn @janis-ax FYI folks, this issue is now fixed in v2.8.2. Thanks again for reporting it.

View full answer

janis-ax · 2023-10-14T22:04:34Z

janis-ax
Oct 14, 2023

I have the same issue! 👍

Terraform used the selected providers to generate the following execution plan. Resource actions are indicated with the following symbols:
  + create

Terraform will perform the following actions:

  # module.kube-hetzner.hcloud_load_balancer.cluster[0] will be created
  + resource "hcloud_load_balancer" "cluster" {
      + delete_protection  = false
      + id                 = (known after apply)
      + ipv4               = (known after apply)
      + ipv6               = (known after apply)
      + labels             = {
          + "cluster"     = "my-k3s"
          + "engine"      = "k3s"
          + "provisioner" = "terraform"
        }
      + load_balancer_type = "lb11"
      + location           = "fsn1"
      + name               = "my-k3s"
      + network_id         = (known after apply)
      + network_ip         = (known after apply)
      + network_zone       = (known after apply)

      + algorithm {
          + type = "round_robin"
        }
    }

  # module.kube-hetzner.null_resource.kustomization will be created
  + resource "null_resource" "kustomization" {
      + id       = (known after apply)
      + triggers = {
          + "helm_values_yaml" = (sensitive value)
          + "options"          = ""
          + "versions"         = <<-EOT
                v1.27
                v1.27.3
                N/A
                N/A
                N/A
                N/A
                v1.14.0
            EOT
        }
    }

Plan: 2 to add, 0 to change, 0 to destroy.

───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────

Saved the plan to: tf.plan

To perform exactly these actions, run the following command to apply:
    terraform apply "tf.plan"
 terraform apply "tf.plan"  
module.kube-hetzner.hcloud_load_balancer.cluster[0]: Creating...
╷
│ Error: name is already used (uniqueness_error)
│ 
│   with module.kube-hetzner.hcloud_load_balancer.cluster[0],
│   on .terraform/modules/kube-hetzner/init.tf line 1, in resource "hcloud_load_balancer" "cluster":
│    1: resource "hcloud_load_balancer" "cluster" {

5 replies

Silvest89 Oct 15, 2023
Collaborator

@mysticaltech
#1001

This is the root issue. While the author did test creating everything from scratch. He didn't account for existing clusters which already have a load balancer with that same name. (HCloud managed)

To fix it either:

Suffix/prefix to the pre-created loadbalancer name (also change it in annotations code piece
Add an enabled attribute for backwards compatibility (eg TF managed, or HCloud managed)

Also:

If Terraform managed deletion protection should be enabled so Hcloud cannot accidently delete it

Anyways @K-MeLeOn @janis-ax you can delete the load balancer in the Hetzner Console and then run terraform apply this should make the loadbalancer terraform managed (it wasn't previously)

mysticaltech Oct 15, 2023
Maintainer

Exactly @Silvest89, thanks for pointing this out. @schlichtanders Do you have the bandwith to publish a fix based on the above recommendations? It would be awesome 🙏

mysticaltech Oct 17, 2023
Maintainer

@Silvest89 @schlichtanders I implemented the fix, at least the start of it. #1034. A review would be appreciated folks .

@Silvest89 I followed your proposed fix. It should let the upgrade go through, but I am not sure that it will replace the old one, probably not, and probably it's a good thing to let people migrate slowly as they see fit. But your opinion would be very much appreciated!

Silvest89 Oct 17, 2023
Collaborator

@mysticaltech
Well the hcloud controller should actually delete the old one if it isn't referenced anymore. Since you have a new name now. This should be a backwards compatible fix. There is one thing missing though I will comment it on the PR

schlichtanders Oct 18, 2023

@Silvest89 and @mysticaltech thank you so much for fixing the pullrequest of mine 🙏 .

mysticaltech · 2023-10-17T22:07:29Z

mysticaltech
Oct 17, 2023
Maintainer

@K-MeLeOn @janis-ax FYI folks, this issue is now fixed in v2.8.2. Thanks again for reporting it.

9 replies

dstockhammer Nov 15, 2023

Sorry I'm not ready to do that to the production cluster, though I do realise that by applying this I may already have passed the point of no return :(

What difference does it make to "the infrastructure" that a different LB already exists? I don't understand that. If it wasn't able to attach the new LB to to the network and assign targets? What I tried instead was the following:

Deleted myname-traefik (that's the new LB that was created just now)
Detached myname (that's the old/existing LB) from the network. At this point the service is down.
Applied Terraform once more; the script recreated myname-traefik same as above; and again didn't attach it to the network and assigned no services. Why? Does existence of the myname LB really make a difference? It's completely unattached to anything Kube-related at this point 🤔

PS - Thanks for the quick reply, I really appreciate it!

dstockhammer Nov 15, 2023

Update: OK I don't know if this is going to lead to any problems down the road, but I decided to import the existing LB instead of creating a new one and having to delete the old one.

> terragrunt state rm 'module.kube-hetzner.hcloud_load_balancer.cluster[0]'
> terragrunt import 'module.kube-hetzner.hcloud_load_balancer.cluster[0]' 0000
> terragrunt apply

Terraform will perform the following actions:

  # module.kube-hetzner.hcloud_load_balancer.cluster[0] will be updated in-place
  ~ resource "hcloud_load_balancer" "cluster" {
        id                 = "0000000"
      ~ labels             = {
          + "cluster"                = "xxx"
          + "engine"                 = "k3s"
          - "hcloud-ccm/service-uid" = "redacted" -> null
          + "provisioner"            = "terraform"
        }
      ~ name               = "xxx" -> "xxx-traefik"
        # (8 unchanged attributes hidden)

        # (4 unchanged blocks hidden)
    }

So far it looks like the old LB has been picked up successfully and everything would have worked without downtime. If everything does indeed continue working, that is...

mysticaltech Nov 17, 2023
Maintainer

@dstockhammer Well done, happy you solved it that way. No reason why it should not work. IMHO.

dstockhammer Nov 17, 2023

One more update; so far it looks like everything is working, except for one thing: New nodes aren't added as targets in the load balancer, regardless of whether they're provisioned manually via a static node pool or by the autoscaler. I solved that issue by manually editing the load balancer to select targets by label. The labels are different for static and autoscaled nodes, but role=agent_node and hcloud/node-group=xxx (the value of this label depends on the configured name of the node group) selects all the workers.

That seems to work fine for me. I'll let you know if I run into any issues in case anyone wants to copy this approach.

mysticaltech Nov 18, 2023
Maintainer

@dstockhammer So it's not playing nice with the Hetzner cloud controller.

K-MeLeOn · 2023-10-18T12:19:00Z

K-MeLeOn
Oct 18, 2023
Author

Nice !
Thank you @Silvest89 and @mysticaltech.
After removing the LB (and change the IP address of the LB domain to the newly created IP address), everything works. 👍

[Version 2.8.2]

0 replies

Lennix · 2024-01-17T11:29:51Z

Lennix
Jan 17, 2024

So I don't know what happened here, but it's still a mess.

I've tried to upgrade multiple times now and I always get stuck at the load balancer. Terraform creates the new load-balancer but it stays disconnected.

I finally could get my old setup working again. Somehow hcloud-cloud-controller-manager had HCLOUD_LOAD_BALANCERS_ENABLED set to false. Was that triggered by the update? Or might that be the cause of the new load-balancer not connecting?

6 replies

Lennix Jan 17, 2024

I removed the LBs countless times. They also were named differently.

K-MeLeOn Jan 17, 2024
Author

Strange behavior, we probably need your kube.tf file (remove sensitive data) to understand.
You don't have a terraform error, right ?

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[2.8.0] - load_balancer name is already used #1028

{{title}}

Replies: 4 comments 20 replies

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

Select a reply

[2.8.0] - load_balancer name is already used #1028

Replies: 4 comments · 20 replies

Silvest89 Oct 15, 2023 Collaborator

mysticaltech Oct 15, 2023 Maintainer

mysticaltech Oct 17, 2023 Maintainer

Silvest89 Oct 17, 2023 Collaborator

mysticaltech Oct 17, 2023 Maintainer

mysticaltech Nov 17, 2023 Maintainer

mysticaltech Nov 18, 2023 Maintainer

K-MeLeOn Oct 18, 2023 Author

K-MeLeOn Jan 17, 2024 Author

Replies: 4 comments 20 replies

Silvest89 Oct 15, 2023
Collaborator

mysticaltech Oct 15, 2023
Maintainer

mysticaltech Oct 17, 2023
Maintainer

Silvest89 Oct 17, 2023
Collaborator

mysticaltech
Oct 17, 2023
Maintainer

mysticaltech Nov 17, 2023
Maintainer

mysticaltech Nov 18, 2023
Maintainer

K-MeLeOn
Oct 18, 2023
Author

K-MeLeOn Jan 17, 2024
Author