Skip destroy if state is empty

[timuthy]

How to categorize this PR?

/area dev-productivity
/area usability
/kind enhancement
/priority normal

What this PR does / why we need it:
If Terraform never ran successfully because of a broken configuration, it'll also fail during the destroy step as described in the linked issue. We can skip further execution in such cases if the state is empty and just remove the finalizer from the state ConfigMap.

Which issue(s) this PR fixes:
Fixes #78

Special notes for your reviewer:
There are other cases in which the state is empty in terms of the Terraform object structure can be found in the state ConfigMap but resources are empty:

  terraform.tfstate: |
    {
      "version": 4,
      "terraform_version": "0.12.29",
      "serial": 62,
      "lineage": "",
      "outputs": {},
      "resources": []
    }

I decided not take a shortcut here because we'd need to explicitly parse the structure and thus introduce a version dependency to Terraform. Still, Terraform will handle such cases gracefully i.e., the only overhead is that we might run Terraform destroy even though it's not necessary (/cc @timebertt).

Along the way I updated:

• Go dependencies
• Go version to 1.16

Release note:

The Terraformer now instantly removes its finalizer from the state `ConfigMap` if the state is empty and `destroy` is called. A separate Terraform `destroy` is not executed.

[timuthy]

/test

[testmachinery]

Testrun: e2e-2wjl5
Workflow: e2e-2wjl5-wf
Phase: Failed

+--------------+--------------+--------+----------+
|     NAME     |     STEP     | PHASE  | DURATION |
+--------------+--------------+--------+----------+
| pod-e2e-test | pod-e2e-test | Failed | 12m2s    |
+--------------+--------------+--------+----------+

[vpnachev]

Thanks!
I have some minor nits.

[timebertt]

Very nice!
/lgtm

[timebertt]

I decided not take a shortcut here because we'd need to explicitly parse the structure and thus introduce a version dependency to Terraform. Still, Terraform will handle such cases gracefully i.e., the only overhead is that we might run Terraform destroy even though it's not necessary (/cc @timebertt).

Sounds reasonable. Great!

[timuthy]

/test

[testmachinery]

Testrun: e2e-gxhrj
Workflow: e2e-gxhrj-wf
Phase: Failed

+--------------+--------------+--------+----------+
|     NAME     |     STEP     | PHASE  | DURATION |
+--------------+--------------+--------+----------+
| pod-e2e-test | pod-e2e-test | Failed | 11m47s   |
+--------------+--------------+--------+----------+

[timebertt]

/test

FYI: the test can only run successfully after the PR build has finished, as it uses the image built by the PR job.

[vpnachev]

/lgtm

[timuthy]

/test

[testmachinery]

Testrun: e2e-k9xpd
Workflow: e2e-k9xpd-wf
Phase: Succeeded

+--------------+--------------+-----------+----------+
|     NAME     |     STEP     |   PHASE   | DURATION |
+--------------+--------------+-----------+----------+
| pod-e2e-test | pod-e2e-test | Succeeded | 2m51s    |
+--------------+--------------+-----------+----------+

[timuthy]

@vpnachev, @timebertt thanks for your reviews. Another commit was necessary for a successful verify execution. PTAL.

[timebertt]

/lgtm