support ipv6 for scp with "none" backend

[phryneas]

Motivation:

Currently, secrets cannot be deployed when deployment.targetHost contains an ipv6.

This is because, while sshing to a ipv6 has the same syntax, scp requires square brackets around the ip. This adds square brackets for calls to scp around any ip with a colon.

I'm no python developer, so please take this PR with a grain of salt. Fortunately, it's not very big ;)

[flokli]

I had this patch applied on top of nixopsUnstable (and it worked deploying when specifying an IPv6 adress), but it seems this breaks copying over secrets:

chekov> copying closure...
network> closures copied successfully
chekov> uploading key ‘foo’...
chekov> error: Traceback (most recent call last):
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 721, in worker
    m.send_keys()
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/backends/__init__.py", line 243, in send_keys
    self.upload_file(tmp, tmp_outfile)
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/backends/__init__.py", line 393, in upload_file
    cmdline += [source, "root@" + self.get_ssh_name(True) + ":" + target]
TypeError: get_ssh_name() takes exactly 1 argument (2 given)

Traceback (most recent call last):
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/bin/..nixops-wrapped-wrapped", line 990, in <module>
    args.op()
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/bin/..nixops-wrapped-wrapped", line 411, in op_deploy
    max_concurrent_activate=args.max_concurrent_activate)
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 1057, in deploy
    self.run_with_notify('deploy', lambda: self._deploy(**kwargs))
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 1046, in run_with_notify
    f()
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 1057, in <lambda>
    self.run_with_notify('deploy', lambda: self._deploy(**kwargs))
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 1013, in _deploy
    dry_activate=dry_activate, max_concurrent_activate=max_concurrent_activate)
  File "/nix/store/vjjpmhn8aqmq0vks1h3s3ck69rxqb57g-nixops-1.6.1pre2706_d5ad09c/lib/python2.7/site-packages/nixops/deployment.py", line 769, in activate_configs
    .format(len(failed), len(res), ", ".join(["‘{0}’".format(x) for x in failed])))
Exception: activation of 1 of 1 machines failed (namely on ‘somehost’)

[phryneas]

Looks like another get_ssh_name is being called - I didn't account for the other backends. Are you definitely using the ǹone` backend?

[flokli]

I observed the regression with the hetzner provider, didn't test deploying secrets with the 'none' provider so far.

[phryneas]

Ah, okay, I need to update the signature to def get_ssh_name(self, scp=False): in the other providers, too, then.

I'll update this PR in the next days. In the meantime, just update def get_ssh_name(self): in the hetzner provider to def get_ssh_name(self, scp=False):, that should solve it for you for now.

[flokli]

In the meantime, just update def get_ssh_name(self): in the hetzner provider to def get_ssh_name(self, scp=False):, that should solve it for you for now.

Looking at the other implementations of get_ssh_name, shouldn't the [%s] wrapping logic better be placed into a helper function in nixops/backends/__init__.py, used by both upload_file and download_file, instead of passing that ssh parameter through all backends, and applying logic there?

[phryneas]

Yup, I had the same thought. I'm quite busy right now, but I'll try to update this PR in the next weeks. It's not like it hasn't been lying around for a while already ;)

[flokli]

Sure, whenever you find the time :-)

[flokli]

@phryneas - poke, could you update this PR?

[phryneas]

@flokli oops, sorry. There you go.

[flokli]

LGTM.

CI is red due to the current nixpkgs breakage with azure-mgmt-compute-4.4.0 (unrelated, fixed by NixOS/nixpkgs#56775).