google_compute_instance: network_interface constantly forces resource recreation

[gmiroshnykov]

I've hit the following issue when trying to migrate from 0.3.7 to 0.4.0.

Basically, google_compute_instance is unusable, because it tries to recreate itself all the time.

Configuration:

resource "google_compute_instance" "test" {
    name = "test-instance"
    machine_type = "n1-standard-1"
    zone = "europe-west1-b"

    disk {
        image = "ubuntu-1404-trusty-v20150316"
        type = "pd-standard"
    }

    network_interface {
        network = "default"
    }
}

output "test" {
    value = "${google_compute_instance.test.name}"
}

First, I run terraform apply:

google_compute_instance.test: Creating...
  can_ip_forward:              "" => "0"
  disk.#:                      "" => "1"
  disk.0.auto_delete:          "" => "1"
  disk.0.image:                "" => "ubuntu-1404-trusty-v20150316"
  disk.0.type:                 "" => "pd-standard"
  machine_type:                "" => "n1-standard-1"
  metadata_fingerprint:        "" => "<computed>"
  name:                        "" => "test-instance"
  network_interface.#:         "" => "1"
  network_interface.0.address: "" => "<computed>"
  network_interface.0.name:    "" => "<computed>"
  network_interface.0.network: "" => "default"
  self_link:                   "" => "<computed>"
  tags_fingerprint:            "" => "<computed>"
  zone:                        "" => "europe-west1-b"
google_compute_instance.test: Creation complete

Apply complete! Resources: 1 added, 0 changed, 0 destroyed.

Next, I run terraform plan with no changes to configuration:

-/+ google_compute_instance.test
    can_ip_forward:              "false" => "0"
    disk.#:                      "1" => "1"
    disk.0.auto_delete:          "true" => "1"
    disk.0.image:                "ubuntu-1404-trusty-v20150316" => "ubuntu-1404-trusty-v20150316"
    disk.0.type:                 "pd-standard" => "pd-standard"
    machine_type:                "n1-standard-1" => "n1-standard-1"
    metadata_fingerprint:        "hDGRxu_A09k=" => "<computed>"
    name:                        "test-instance" => "test-instance"
    network_interface.#:         "1" => "1"
    network_interface.0.address: "10.240.89.254" => "<computed>"
    network_interface.0.name:    "nic0" => "<computed>"
    network_interface.0.network: "https://www.googleapis.com/compute/v1/projects/my-project-name/global/networks/default" => "default" (forces new resource)
    self_link:                   "https://www.googleapis.com/compute/v1/projects/my-project-name/zones/europe-west1-b/instances/test-instance" => "<computed>"
    tags_fingerprint:            "42WmSpB8rSM=" => "<computed>"
    zone:                        "europe-west1-b" => "europe-west1-b"

As you can see, Terraform insists on recreating instance because network_interface.0.network value is different.

[gmiroshnykov]

Might be related to #1286.

[MerlinDMC]

I was just going to post the same issue after hitting that yesterday as well.
It seems to persist the self_link instead of the network name in the state but requires the actual network name for creation.

I did help me around that for now by editing the terraform config once the resource got created:

  network_interface {
    // network = "${google_compute_network.default.name}"
    network = "${google_compute_network.default.self_link}"
    access_config {
      // Ephemeral IP
    }
  }

which is a stupid way around this but getting me working again.

I'm currently looking for a way to turn the Network self_link into a usable name via the google api for submitting a patch - but so far I'm not that lucky.

Maybe the provider tests should have a case for testing the plan resolution after a resource got created and test for the number of needed transformations to be zero.

[MerlinDMC]

This does fix my current problems in a first stupid way but I doubt it will fix the complete problem:

diff --git a/builtin/providers/google/resource_compute_instance.go b/builtin/providers/google/resource_compute_instance.go
index d89e82a..40a7fe6 100644
--- a/builtin/providers/google/resource_compute_instance.go
+++ b/builtin/providers/google/resource_compute_instance.go
@@ -3,6 +3,7 @@ package google
 import (
        "fmt"
        "log"
+       "path"
        "time"

        "github.com/hashicorp/terraform/helper/hashcode"
@@ -572,7 +573,7 @@ func resourceComputeInstanceRead(d *schema.ResourceData, meta interface{}) error
                        networkInterfaces = append(networkInterfaces, map[string]interface{}{
                                "name":          iface.Name,
                                "address":       iface.NetworkIP,
-                               "network":       iface.Network,
+                               "network":       path.Base(iface.Network),
                                "access_config": accessConfigs,
                        })
                }

[thijsschnitger]

I've hit this bug as well, using the workaround proposed by @MerlinDMC for now.

[phinze]

Hey folks, just merged @dainis's fix into master. If that doesn't solve this issue, let me know and we can re-evaluate. 👍

[thijsschnitger]

worksforme thanks!

[gmiroshnykov]

I can also confirm that this works, thanks a lot!

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.