path: root/overlays/nixops/hetzner_cloud.patch

From 272e50d0b0262e49cdcaad42cdab57aad183d1c2 Mon Sep 17 00:00:00 2001
From: goodraven
 <employee-pseudonym-7f597def-7eeb-47f8-b10a-0724f2ba59a9@google.com>
Date: Thu, 3 May 2018 22:24:58 -0700
Subject: [PATCH] Initial commit adding support for hetzner cloud

This is based on the digital ocean backend. It also uses nixos-infect. I extended nixos-infect to be generic
for both backends.

Fixes #855
---
 examples/trivial-hetzner-cloud.nix |  12 ++
 nix/eval-machine-info.nix          |   1 +
 nix/hetzner-cloud.nix              |  56 +++++++
 nix/options.nix                    |   1 +
 nixops/backends/hetzner_cloud.py   | 230 +++++++++++++++++++++++++++++
 nixops/data/nixos-infect           |  77 +++++++---
 6 files changed, 354 insertions(+), 23 deletions(-)
 create mode 100644 examples/trivial-hetzner-cloud.nix
 create mode 100644 nix/hetzner-cloud.nix
 create mode 100644 nixops/backends/hetzner_cloud.py

diff --git a/examples/trivial-hetzner-cloud.nix b/examples/trivial-hetzner-cloud.nix
new file mode 100644
index 000000000..c61add6bb
--- /dev/null
+++ b/examples/trivial-hetzner-cloud.nix
@@ -0,0 +1,12 @@
+{
+  resources.sshKeyPairs.ssh-key = {};
+
+  machine = { config, pkgs, ... }: {
+    services.openssh.enable = true;
+
+    deployment.targetEnv = "hetznerCloud";
+    deployment.hetznerCloud.serverType = "cx11";
+
+    networking.firewall.allowedTCPPorts = [ 22 ];
+  };
+}
diff --git a/nix/eval-machine-info.nix b/nix/eval-machine-info.nix
index 2884b4b47..6a7205786 100644
--- a/nix/eval-machine-info.nix
+++ b/nix/eval-machine-info.nix
@@ -309,6 +309,7 @@ rec {
           digitalOcean = optionalAttrs (v.config.deployment.targetEnv == "digitalOcean") v.config.deployment.digitalOcean;
           gce = optionalAttrs (v.config.deployment.targetEnv == "gce") v.config.deployment.gce;
           hetzner = optionalAttrs (v.config.deployment.targetEnv == "hetzner") v.config.deployment.hetzner;
+          hetznerCloud = optionalAttrs (v.config.deployment.targetEnv == "hetznerCloud") v.config.deployment.hetznerCloud;
           container = optionalAttrs (v.config.deployment.targetEnv == "container") v.config.deployment.container;
           route53 = v.config.deployment.route53;
           virtualbox =
diff --git a/nix/hetzner-cloud.nix b/nix/hetzner-cloud.nix
new file mode 100644
index 000000000..21d148c1a
--- /dev/null
+++ b/nix/hetzner-cloud.nix
@@ -0,0 +1,56 @@
+{ config, pkgs, lib, utils, ... }:
+
+with utils;
+with lib;
+with import ./lib.nix lib;
+
+let
+  cfg = config.deployment.hetznerCloud;
+in
+{
+  ###### interface
+  options = {
+
+    deployment.hetznerCloud.authToken = mkOption {
+      default = "";
+      example = "8b2f4e96af3997853bfd4cd8998958eab871d9614e35d63fab45a5ddf981c4da";
+      type = types.str;
+      description = ''
+        The API auth token. We're checking the environment for
+        <envar>HETZNER_CLOUD_AUTH_TOKEN</envar> first and if that is
+        not set we try this auth token.
+      '';
+    };
+
+    deployment.hetznerCloud.datacenter = mkOption {
+      example = "fsn1-dc8";
+      default = null;
+      type = types.nullOr types.str;
+      description = ''
+        The datacenter.
+      '';
+    };
+
+    deployment.hetznerCloud.location = mkOption {
+      example = "fsn1";
+      default = null;
+      type = types.nullOr types.str;
+      description = ''
+        The location.
+      '';
+    };
+
+    deployment.hetznerCloud.serverType = mkOption {
+      example = "cx11";
+      type = types.str;
+      description = ''
+        Name or id of server types.
+      '';
+    };
+  };
+
+  config = mkIf (config.deployment.targetEnv == "hetznerCloud") {
+    nixpkgs.system = mkOverride 900 "x86_64-linux";
+    services.openssh.enable = true;
+  };
+}
diff --git a/nix/options.nix b/nix/options.nix
index 0866c3ab8..db021f74d 100644
--- a/nix/options.nix
+++ b/nix/options.nix
@@ -22,6 +22,7 @@ in
       ./keys.nix
       ./gce.nix
       ./hetzner.nix
+      ./hetzner-cloud.nix
       ./container.nix
       ./libvirtd.nix
     ];
diff --git a/nixops/backends/hetzner_cloud.py b/nixops/backends/hetzner_cloud.py
new file mode 100644
index 000000000..a2cb176b9
--- /dev/null
+++ b/nixops/backends/hetzner_cloud.py
@@ -0,0 +1,230 @@
+# -*- coding: utf-8 -*-
+"""
+A backend for hetzner cloud.
+
+This backend uses nixos-infect (which uses nixos LUSTRATE) to infect a
+hetzner cloud instance. The setup requires two reboots, one for
+the infect itself, another after we pushed the nixos image.
+"""
+import os
+import os.path
+import time
+import socket
+
+import requests
+
+import nixops.resources
+from nixops.backends import MachineDefinition, MachineState
+from nixops.nix_expr import Function, RawValue
+import nixops.util
+import nixops.known_hosts
+
+infect_path = os.path.abspath(os.path.join(os.path.dirname(__file__), '..', 'data', 'nixos-infect'))
+
+API_HOST = 'api.hetzner.cloud'
+
+class ApiError(Exception):
+    pass
+
+class ApiNotFoundError(ApiError):
+    pass
+
+class HetznerCloudDefinition(MachineDefinition):
+    @classmethod
+    def get_type(cls):
+        return "hetznerCloud"
+
+    def __init__(self, xml, config):
+        MachineDefinition.__init__(self, xml, config)
+        self.auth_token = config["hetznerCloud"]["authToken"]
+        self.location = config["hetznerCloud"]["location"]
+        self.datacenter = config["hetznerCloud"]["datacenter"]
+        self.server_type = config["hetznerCloud"]["serverType"]
+
+    def show_type(self):
+        return "{0} [{1}]".format(self.get_type(), self.location or self.datacenter or 'any location')
+
+
+class HetznerCloudState(MachineState):
+    @classmethod
+    def get_type(cls):
+        return "hetznerCloud"
+
+    state = nixops.util.attr_property("state", MachineState.MISSING, int)  # override
+    public_ipv4 = nixops.util.attr_property("publicIpv4", None)
+    public_ipv6 = nixops.util.attr_property("publicIpv6", None)
+    location = nixops.util.attr_property("hetznerCloud.location", None)
+    datacenter = nixops.util.attr_property("hetznerCloud.datacenter", None)
+    server_type = nixops.util.attr_property("hetznerCloud.serverType", None)
+    auth_token = nixops.util.attr_property("hetznerCloud.authToken", None)
+    server_id = nixops.util.attr_property("hetznerCloud.serverId", None, int)
+
+    def __init__(self, depl, name, id):
+        MachineState.__init__(self, depl, name, id)
+        self.name = name
+
+    def get_ssh_name(self):
+        return self.public_ipv4
+
+    def get_ssh_flags(self, *args, **kwargs):
+        super_flags = super(HetznerCloudState, self).get_ssh_flags(*args, **kwargs)
+        return super_flags + [
+            '-o', 'UserKnownHostsFile=/dev/null',
+            '-o', 'StrictHostKeyChecking=no',
+            '-i', self.get_ssh_private_key_file(),
+        ]
+
+    def get_physical_spec(self):
+        return Function("{ ... }", {
+            'imports': [ RawValue('<nixpkgs/nixos/modules/profiles/qemu-guest.nix>') ],
+            ('boot', 'loader', 'grub', 'device'): 'nodev',
+            ('fileSystems', '/'): { 'device': '/dev/sda1', 'fsType': 'ext4'},
+            ('users', 'extraUsers', 'root', 'openssh', 'authorizedKeys', 'keys'): [self.depl.active_resources.get('ssh-key').public_key],
+        })
+
+    def get_ssh_private_key_file(self):
+        return self.write_ssh_private_key(self.depl.active_resources.get('ssh-key').private_key)
+
+    def create_after(self, resources, defn):
+        # make sure the ssh key exists before we do anything else
+        return {
+            r for r in resources if
+            isinstance(r, nixops.resources.ssh_keypair.SSHKeyPairState)
+        }
+
+    def get_auth_token(self):
+        return os.environ.get('HETZNER_CLOUD_AUTH_TOKEN', self.auth_token)
+
+    def _api(self, path, method=None, data=None, json=True):
+        """Basic wrapper around requests that handles auth and serialization."""
+        assert path[0] == '/'
+        url = 'https://%s%s' % (API_HOST, path)
+        token = self.get_auth_token()
+        if not token:
+            raise Exception('No hetzner cloud auth token set')
+        headers = {
+            'Authorization': 'Bearer '+self.get_auth_token(),
+        }
+        res = requests.request(
+            method=method,
+            url=url,
+            json=data,
+            headers=headers)
+
+        if res.status_code == 404:
+            raise ApiNotFoundError('Not Found: %r' % path)
+        elif not res.ok:
+            raise ApiError('Response for %s %s has status code %d: %s' % (method, path, res.status_code, res.content))
+        if not json:
+            return
+        try:
+            res_data = res.json()
+        except ValueError as e:
+            raise ApiError('Response for %s %s has invalid JSON (%s): %r' % (method, path, e, res.content))
+        return res_data
+
+
+    def destroy(self, wipe=False):
+        if not self.server_id:
+            self.log('server {} was never made'.format(self.name))
+            return
+        self.log('destroying server {} with id {}'.format(self.name, self.server_id))
+        try:
+            res = self._api('/v1/servers/%s' % (self.server_id), method='DELETE')
+        except ApiNotFoundError:
+            self.log("server not found - assuming it's been destroyed already")
+
+        self.public_ipv4 = None
+        self.server_id = None
+
+        return True
+
+    def _create_ssh_key(self, public_key):
+        """Create or get an ssh key and return an id."""
+        public_key = public_key.strip()
+        res = self._api('/v1/ssh_keys', method='GET')
+        name = 'nixops-%s-%s' % (self.depl.uuid, self.name)
+        deletes = []
+        for key in res['ssh_keys']:
+            if key['public_key'].strip() == public_key:
+                return key['id']
+            if key['name'] == name:
+                deletes.append(key['id'])
+        for d in deletes:
+            # This reply is empty, so don't decode json.
+            self._api('/v1/ssh_keys/%d' % d, method='DELETE', json=False)
+        res = self._api('/v1/ssh_keys', method='POST', data={
+            'name': name,
+            'public_key': public_key,
+        })
+        return res['ssh_key']['id']
+
+    def create(self, defn, check, allow_reboot, allow_recreate):
+        ssh_key = self.depl.active_resources.get('ssh-key')
+        if ssh_key is None:
+            raise Exception('Please specify a ssh-key resource (resources.sshKeyPairs.ssh-key = {}).')
+
+        self.set_common_state(defn)
+
+        if self.server_id is not None:
+            return
+
+        ssh_key_id = self._create_ssh_key(ssh_key.public_key)
+
+        req = {
+                'name': self.name,
+                'server_type': defn.server_type,
+                'start_after_create': True,
+                'image': 'debian-9',
+                'ssh_keys': [
+                    ssh_key_id,
+                ],
+        }
+
+        if defn.datacenter:
+            req['datacenter'] = defn.datacenter
+        elif defn.location:
+            req['location'] = defn.location
+
+        self.log_start("creating server ...")
+        create_res = self._api('/v1/servers', method='POST', data=req)
+        self.server_id = create_res['server']['id']
+        self.public_ipv4 = create_res['server']['public_net']['ipv4']['ip']
+        self.public_ipv6 = create_res['server']['public_net']['ipv6']['ip']
+        self.datacenter = create_res['server']['datacenter']['name']
+        self.location = create_res['server']['datacenter']['location']['name']
+
+        action = create_res['action']
+        action_path = '/v1/servers/%d/actions/%d' % (self.server_id, action['id'])
+
+        while action['status'] == 'running':
+            time.sleep(1)
+            res = self._api(action_path, method='GET')
+            action = res['action']
+
+        if action['status'] != 'success':
+            raise Exception('unexpected status: %s' % action['status'])
+
+        self.log_end("{}".format(self.public_ipv4))
+
+        self.wait_for_ssh()
+        self.log_start("running nixos-infect")
+        self.run_command('bash </dev/stdin 2>&1', stdin=open(infect_path))
+        self.reboot_sync()
+
+    def reboot(self, hard=False):
+        if hard:
+            self.log("sending hard reset to server...")
+            res = self._api('/v1/servers/%d/actions/reset' % self.server_id, method='POST')
+            action = res['action']
+            action_path = '/v1/servers/%d/actions/%d' % (self.server_id, action['id'])
+            while action['status'] == 'running':
+                time.sleep(1)
+                res = self._api(action_path, method='GET')
+                action = res['action']
+            if action['status'] != 'success':
+                raise Exception('unexpected status: %s' % action['status'])
+            self.wait_for_ssh()
+            self.state = self.STARTING
+        else:
+            MachineState.reboot(self, hard=hard)
diff --git a/nixops/data/nixos-infect b/nixops/data/nixos-infect
index 66634357b..437a2ec61 100644
--- a/nixops/data/nixos-infect
+++ b/nixops/data/nixos-infect
@@ -68,26 +68,49 @@ makeConf() {
 }
 EOF
   # (nixos-generate-config will add qemu-user and bind-mounts, so avoid)
+  local disk
+  if [ -e /dev/sda ]; then
+    disk=/dev/sda
+  else
+    disk=/dev/vda
+  fi
   cat > /etc/nixos/hardware-configuration.nix << EOF
 { ... }:
 {
   imports = [ <nixpkgs/nixos/modules/profiles/qemu-guest.nix> ];
-  boot.loader.grub.device = "/dev/vda";
-  fileSystems."/" = { device = "/dev/vda1"; fsType = "ext4"; };
+  boot.loader.grub.device = "${disk}";
+  fileSystems."/" = { device = "${disk}1"; fsType = "ext4"; };
 }
 EOF
 
   local IFS=$'\n'
-  ens3_ip4s=($(ip address show dev eth0 | grep 'inet '          | sed -r 's|.*inet ([0-9.]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
-  ens3_ip6s=($(ip address show dev eth0 | grep 'inet6 .*global' | sed -r 's|.*inet6 ([0-9a-f:]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
-  ens4_ip4s=($(ip address show dev eth1 | grep 'inet '          | sed -r 's|.*inet ([0-9.]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
-  ens4_ip6s=($(ip address show dev eth1 | grep 'inet6 .*global' | sed -r 's|.*inet6 ([0-9a-f:]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
-  gateway=($(ip route show dev eth0 | grep default | sed -r 's|default via ([0-9.]+).*|\1|'))
-  gateway6=($(ip -6 route show dev eth0 | grep default | sed -r 's|default via ([0-9a-f:]+).*|\1|'))
-  ether0=($(ip address show dev eth0 | grep link/ether | sed -r 's|.*link/ether ([0-9a-f:]+) .*|\1|'))
-  ether1=($(ip address show dev eth1 | grep link/ether | sed -r 's|.*link/ether ([0-9a-f:]+) .*|\1|'))
+  gateway=($(ip route show | grep default | sed -r 's|default via ([0-9.]+).*|\1|'))
+  gateway6=($(ip -6 route show | grep default | sed -r 's|default via ([0-9a-f:]+).*|\1|'))
+  interfaces=($(ip link | awk -F ': ' '/^[0-9]*: / {if ($2 != "lo") {print $2}}'))
   nameservers=($(grep ^nameserver /etc/resolv.conf | cut -f2 -d' '))
 
+  # Predict the predictable name for each interface since that is enabled in
+  # the nixos system.
+  declare -A predictable_names
+  for interface in ${interfaces[@]}; do
+    # udevadm prints out the candidate names which will be selected if
+    # available in this order.
+    local name=$(udevadm info /sys/class/net/$interface | awk -F = '
+    /^E: ID_NET_NAME_FROM_DATABASE=/ {arr[1]=$2}
+    /^E: ID_NET_NAME_ONBOARD=/ {arr[2]=$2}
+    /^E: ID_NET_NAME_SLOT=/ {arr[3]=$2}
+    /^E: ID_NET_NAME_PATH=/ {arr[4]=$2}
+    /^E: ID_NET_NAME_MAC=/ {arr[5]=$2}
+    END {for (i=1;i<6;i++) {if (length(arr[i]) > 0) { print arr[i]; break}}}')
+    if [ -z "$name" ]; then
+      echo Could not determine predictable name for interface $interface
+    fi
+    predictable_names[$interface]=$name
+  done
+
+  # Take a gamble on the first interface being able to reach the gateway.
+  local default_interface=${predictable_names[${interfaces[0]}]}
+
   cat > /etc/nixos/networking.nix << EOF
 { ... }: {
   # This file was populated at runtime with the networking
@@ -96,25 +119,27 @@ EOF
     nameservers = [$(for a in ${nameservers[@]}; do echo -n "
       \"$a\""; done)
     ];
-    defaultGateway = "${gateway}";
-    defaultGateway6 = "${gateway6}";
+    defaultGateway = {address = "${gateway}"; interface = "${default_interface}";};
+    defaultGateway6 = {address = "${gateway6}"; interface = "${default_interface}";};
     interfaces = {
-      ens3 = {
-        ip4 = [$(for a in ${ens3_ip4s[@]}; do echo -n "
-          $a"; done)
-        ];
-        ip6 = [$(for a in ${ens3_ip6s[@]}; do echo -n "
-          $a"; done)
-        ];
-      };
-      ens4 = {
-        ip4 = [$(for a in ${ens4_ip4s[@]}; do echo -n "
+EOF
+
+  for interface in ${interfaces[@]}; do
+    ip4s=($(ip address show dev $interface | grep 'inet '          | sed -r 's|.*inet ([0-9.]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
+    ip6s=($(ip address show dev $interface | grep 'inet6 .*global' | sed -r 's|.*inet6 ([0-9a-f:]+)/([0-9]+).*|{ address="\1"; prefixLength=\2; }|'))
+    cat >> /etc/nixos/networking.nix << EOF
+      ${predictable_names[$interface]} = {
+        ip4 = [$(for a in ${ip4s[@]}; do echo -n "
           $a"; done)
         ];
-        ip6 = [$(for a in ${ens4_ip6s[@]}; do echo -n "
+        ip6 = [$(for a in ${ip6s[@]}; do echo -n "
           $a"; done)
         ];
       };
+EOF
+  done
+
+  cat >> /etc/nixos/networking.nix << EOF
     };
   };
 }
@@ -154,6 +179,12 @@ export HOME="/root"
 groupadd -r nixbld -g 30000
 seq 1 10 | xargs -I{} useradd -c "Nix build user {}" -d /var/empty -g nixbld -G nixbld -M -N -r -s `which nologin` nixbld{}
 
+if ! which curl >/dev/null 2>/dev/null; then
+  if which apt-get >/dev/null 2>/dev/null; then
+    apt-get update && apt-get install -y curl
+  fi
+fi
+
 curl https://nixos.org/nix/install | sh
 
 source ~/.nix-profile/etc/profile.d/nix.sh