Nebula

Nebula は、トンネリングと暗号化を使用して参加ホスト間で安全なプライベートメッシュネットワークを作成する、ユーザー空間メッシュ仮想プライベートネットワーク（VPN）デーモンです。

インストール

基本概念と用語

Nebula は、tinc に触発されたメッシュ VPN 技術です。メッシュ VPN では、個々のノードが互いに直接トンネルを形成します。これにより、中央ノードを経由せずに、ノード間で高速な直接通信を可能にします。ノードは、証明書機関によって署名された証明書を使用して認証されます。

これは、ピアツーピア VPN 技術である WireGuard とは対照的です（ただし、WireGuard 用のメッシュネットワークマネージャーが存在します。例：innernet および wesher^AUR）。

これは、星形トポロジー（ハブアンドスポークとも呼ばれる）を使用する OpenVPN とも異なります。

証明書機関: 証明書機関は、それに署名することでホスト証明書を作成します。
ライトハウス: Nebula ネットワークでは、通常、他のノードの情報ハブとして機能する少なくとも1つのライトハウスノードがあります。ライトハウスノードは、他のノードが互いを見つけてネットワークメッシュを形成するのを助けます。
ノード: Nebula ネットワークのノード。
Nebula IP: Nebula ネットワーク内のノードの IP アドレス。VPN IP としても知られています。
ルーティング可能な IP: ノードの「通常の」または「ネイティブ」の IP アドレス。これは、ノードの場所とネットワークの設定方法によって、公開 IP アドレスまたはプライベート IP アドレスのいずれかになります。ノードは複数のルーティング可能な IP アドレスを持つことができます。

Example: Simple mesh VPN

Network setup

In this example, we have 3 nodes:

lighthouse
- Nebula IP: 192.168.100.1
- Routable IP: 12.34.56.78
hostA
- Nebula IP: 192.168.100.101
- Routable IP: 10.0.0.22
hostB
- Nebula IP: 192.168.100.102
- Routable IP: 23.45.67.89

The lighthouse has a public static IP address and is reachable by hostA and hostB. hostA lives behind a NAT. hostB has a public IP address.

In our case, we will use a /24 subnet for the VPN network. We will call this network "My Nebula Network".

Certificate and key generation

First, generate the CA certificate and private key with nebula-cert ca -name "My Nebula Network". This will create two files:

ca.crt: The CA certificate file
ca.key: the CA private key

Subsequently, generate the certificate and private key files for the nodes in the network:

$ nebula-cert sign -name lighthouse -ip 192.168.100.1/24
$ nebula-cert sign -name hostA -ip 192.168.100.101/24
$ nebula-cert sign -name hostB -ip 192.168.100.102/24

Notice that we did not specify ca.crt and ca.key. By default, nebula-cert looks for those files in the current directory.

After this step, we will have these files:

lighthouse.crt, lighthouse.key
hostA.crt, hostA.key
hostB.crt, hostB.key

Configuration

Create this configuration file on the lighthouse node:

/etc/nebula/config.yml

pki:
  ca: /etc/nebula/ca.crt
  cert: /etc/nebula/lighthouse.crt
  key: /etc/nebula/lighthouse.key

lighthouse:
  am_lighthouse: true

firewall:
  outbound:
    - port: any
      proto: any
      host: any
  inbound:
    - port: any
      proto: any
      host: any

Create this configuration file on hostA:

/etc/nebula/config.yml

pki:
  ca: /etc/nebula/ca.crt
  cert: /etc/nebula/hostA.crt
  key: /etc/nebula/hostA.key

static_host_map:
  "192.168.100.1": ["12.34.56.78:4242"]

lighthouse:
  hosts:
    - "192.168.100.1"

punchy:
  punch: true

firewall:
  outbound:
    - port: any
      proto: any
      host: any
  inbound:
    - port: any
      proto: any
      host: any

Finally, use this configuration file for hostB:

/etc/nebula/config.yml

pki:
  ca: /etc/nebula/ca.crt
  cert: /etc/nebula/hostB.crt
  key: /etc/nebula/hostB.key

static_host_map:
  "192.168.100.1": ["12.34.56.78:4242"]

lighthouse:
  hosts:
    - "192.168.100.1"

firewall:
  outbound:
    - port: any
      proto: any
      host: any
  inbound:
    - port: any
      proto: any
      host: any

Distribute certificates and private keys

Because the certificates and private keys were generated by the certificate authority, they need to be distributed to each node. SCP and SFTP are suitable for this purpose.

Specifically:

ca.crt should be copied to all 3 nodes: lighthouse, hostA, and hostB
lighthouse.crt and lighthouse.key should be copied to the lighthouse node
hostA.crt and hostA.key should be copied to hostA
hostB.crt and hostB.key should be copied to hostB

ノート: The ca.key file does not have to be copied over to any node. Keep it safe (do not lose it) and secure (do not leak it).

Start the nebula daemon

On each node, start nebula.service. Optionally, enable it so that it will be started on boot.

Note that it does not matter which node starts the nebula daemon. The lighthouse node can even be started last. Each individual node always tries to connect to the list of known lighthouse nodes, so any network interruption can be rectified quickly.

Test for mesh functionality

With a mesh network, every node is directly connected to every other node. So, even if the connection between lighthouse and both hostA and hostB is slow, traffic between hostA and hostB can be fast, as long as there is a direct link between those two.

This can be demonstrated by a simple ping test on hostA:

$ ping -c 5 12.34.56.78

PING 12.34.56.78 (12.34.56.78) 56(84) bytes of data.
64 bytes from 12.34.56.78: icmp_seq=1 ttl=56 time=457 ms
64 bytes from 12.34.56.78: icmp_seq=2 ttl=56 time=480 ms
64 bytes from 12.34.56.78: icmp_seq=3 ttl=56 time=262 ms
64 bytes from 12.34.56.78: icmp_seq=4 ttl=56 time=199 ms
64 bytes from 12.34.56.78: icmp_seq=5 ttl=56 time=344 ms

--- 12.34.56.78 ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4004ms
rtt min/avg/max/mdev = 199.141/348.555/480.349/108.654 ms

$ ping -c 5 192.168.100.1

PING 192.168.100.1 (192.168.100.1) 56(84) bytes of data.
64 bytes from 192.168.100.1: icmp_seq=1 ttl=64 time=218 ms
64 bytes from 192.168.100.1: icmp_seq=2 ttl=64 time=241 ms
64 bytes from 192.168.100.1: icmp_seq=3 ttl=64 time=264 ms
64 bytes from 192.168.100.1: icmp_seq=4 ttl=64 time=288 ms
64 bytes from 192.168.100.1: icmp_seq=5 ttl=64 time=163 ms

--- 192.168.100.1 ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4004ms
rtt min/avg/max/mdev = 162.776/234.874/288.073/42.902 ms

$ ping -c 5 192.168.100.102

PING 192.168.100.102 (192.168.100.102) 56(84) bytes of data.
64 bytes from 192.168.100.102: icmp_seq=1 ttl=64 time=106 ms
64 bytes from 192.168.100.102: icmp_seq=2 ttl=64 time=2.14 ms
64 bytes from 192.168.100.102: icmp_seq=3 ttl=64 time=4.53 ms
64 bytes from 192.168.100.102: icmp_seq=4 ttl=64 time=4.29 ms
64 bytes from 192.168.100.102: icmp_seq=5 ttl=64 time=5.39 ms

--- 192.168.100.102 ping statistics ---
5 packets transmitted, 5 received, 0% packet loss, time 4006ms
rtt min/avg/max/mdev = 2.136/24.535/106.344/40.918 ms

Notice that the connection between hostA and lighthouse is slow, but the connection between hostA and hostB is very fast. Also notice that the first packet between hostA and hostB is delayed a bit, but subsequent packets take almost no time at all.

Configuration options

listen.port: This is the listening port for the nebula daemon, which by default is 4242. On a lighthouse node, or a node with a static IP address, set this to any other number in order to personalize your setup and reduce the chances of unwanted service discovery and DDoS attacks on that port. Then update static_host_map to reflect the change.; On a node with a dynamic IP address, it is recommended to set this to 0, such the nebula daemon will use a random port for communication.
logging.level: By default, the nebula daemon logs INFO-level messages. Thus handshakes are printed, and this can generate a lot of log messages. Set it to warning in order to reduce the amount of messages logged.
relay: This option can be used if a node cannot be reached directly from another node. Relay nodes help forward the communication between such nodes.
firewall: This option can be used to allow only certain traffic to and from a node.

Troubleshooting

My lighthouse node takes forever to handshake

If your lighthouse node needs a long time to handshake, and it prints multiple handshake messages all at once when handshake is completed, maybe it does not support recvmmsg(). To get around this issue, add this configuration option:

/etc/nebula/config.yml

listen:
  batch: 1

This problem usually happens if your Linux kernel is too old (<2.6.34). The proper solution is to upgrade it.