Terraform destroy for cloudscale.ch fails at the first time

[ccremer]

With appuio/openshift4-docs#96 implemented, terraform destroy takes 2 times to complete.
The first time the error is something like

module.cluster.cloudscale_subnet.privnet_subnet: Destroying... [id=f0c8cbf0-5d5e-48b8-9379-140cf19f7330]
module.cluster.cloudscale_server.lb[0]: Destroying... [id=6e8f4c10-00f3-45c8-99f2-67ffbd772ec6]
module.cluster.cloudscale_server.lb[1]: Destroying... [id=c42a7f61-eef3-4932-bc0c-9bb69d5b3d62]
module.cluster.cloudscale_server.lb[1]: Destruction complete after 7s
module.cluster.cloudscale_server.lb[0]: Still destroying... [id=6e8f4c10-00f3-45c8-99f2-67ffbd772ec6, 10s elapsed]
module.cluster.cloudscale_server.lb[0]: Destruction complete after 10s
module.cluster.cloudscale_server_group.lb[0]: Destroying... [id=c39aa9ae-5aaa-43ca-b6a2-9b99f992eedf]
module.cluster.null_resource.register_lb[1]: Destroying... [id=3075016582769078992]
module.cluster.null_resource.register_lb[0]: Destroying... [id=357206802485481663]
module.cluster.local_file.lb_hieradata[0]: Destroying... [id=643ab3342838dac0aaca5a9d30a947f9d9ea52a1]
module.cluster.null_resource.register_lb[0]: Destruction complete after 0s
module.cluster.null_resource.register_lb[1]: Destruction complete after 0s
module.cluster.local_file.lb_hieradata[0]: Destruction complete after 0s
module.cluster.random_id.lb[0]: Destroying... [id=hg]
module.cluster.cloudscale_floating_ip.nat_vip[0]: Destroying... [id=5.102.151.35]
module.cluster.random_id.lb[1]: Destroying... [id=HA]
module.cluster.gitfile_checkout.appuio_hieradata[0]: Destroying... [id=./appuio_hieradata]
module.cluster.cloudscale_floating_ip.router_vip[0]: Destroying... [id=5.102.150.209]
module.cluster.random_id.lb[0]: Destruction complete after 0s
module.cluster.random_id.lb[1]: Destruction complete after 0s
module.cluster.cloudscale_floating_ip.api_vip[0]: Destroying... [id=5.102.151.109]
module.cluster.cloudscale_server_group.lb[0]: Destruction complete after 2s
module.cluster.cloudscale_floating_ip.router_vip[0]: Destruction complete after 3s
module.cluster.cloudscale_floating_ip.api_vip[0]: Destruction complete after 4s
module.cluster.cloudscale_floating_ip.nat_vip[0]: Destruction complete after 4s

Error: Error while running git pull --ff-only origin: exit status 128
Working dir: ./appuio_hieradata
Output: From https://git.vshn.net/appuio/appuio_hieradata
   6e37293..7f44eca  master     -> origin/master
fatal: Not possible to fast-forward, aborting.

Error: Error deleting subnet f0c8cbf0-5d5e-48b8-9379-140cf19f7330: detail: There are still one or more interfaces in this subnet.

After running terraform destroy again, it wants to delete the following resources:

Terraform will perform the following actions:

  # module.cluster.cloudscale_network.privnet will be destroyed
  - resource "cloudscale_network" "privnet" {
      - auto_create_ipv4_subnet = false -> null
      - href                    = "https://api.cloudscale.ch/v1/networks/9bf5f692-3a2a-429a-9caa-c3a2a83939d8" -> null
      - id                      = "9bf5f692-3a2a-429a-9caa-c3a2a83939d8" -> null
      - mtu                     = 9000 -> null
      - name                    = "privnet-c-falling-shadow-3833" -> null
      - subnets                 = [] -> null
      - zone_slug               = "rma1" -> null
    }

  # module.cluster.cloudscale_subnet.privnet_subnet will be destroyed
  - resource "cloudscale_subnet" "privnet_subnet" {
      - cidr            = "172.18.200.0/24" -> null
      - dns_servers     = [
          - "5.102.144.101",
          - "5.102.144.102",
        ] -> null
      - gateway_address = "172.18.200.1" -> null
      - href            = "https://api.cloudscale.ch/v1/subnets/f0c8cbf0-5d5e-48b8-9379-140cf19f7330" -> null
      - id              = "f0c8cbf0-5d5e-48b8-9379-140cf19f7330" -> null
      - network_href    = "https://api.cloudscale.ch/v1/networks/9bf5f692-3a2a-429a-9caa-c3a2a83939d8" -> null
      - network_name    = "privnet-c-falling-shadow-3833" -> null
      - network_uuid    = "9bf5f692-3a2a-429a-9caa-c3a2a83939d8" -> null
    }

  # module.cluster.gitfile_checkout.appuio_hieradata[0] will be destroyed
  - resource "gitfile_checkout" "appuio_hieradata" {
      - branch = "tf/lbaas/c-falling-shadow-3833" -> null
      - head   = "6e372938eb594ccd56988ca14866db4f8f74b507" -> null
      - id     = "./appuio_hieradata" -> null
      - path   = "./appuio_hieradata" -> null
      - repo   = "https://project_368_bot@git.vshn.net/appuio/appuio_hieradata.git" -> null
    }

Plan: 0 to add, 0 to change, 3 to destroy.

I suspect it seems it fails to destroy a private subnet if a VM is also being destroyed right in the same time.

Workaround

Invoke terraform destroy 2 times, that successfully removes all resources.