In the previous post we installed Terraform and used it to create a VPC in AWS , subnets, instances and all the necessary elements for a lab with Wireguard.

In this post we’ll install and configure a Wireguard server, a remote client and use it to access an EC2 instance in a private subnet.


The EC2 instance we installed to host Wireguard has a public IP address.

We can get its IP using AWS CLI

aws ec2 describe-network-interfaces --query NetworkInterfaces[*].PrivateIpAddresses[*].Association.PublicIp

I described how to install aws cli in my previous post

If AWS CLI is not available, we can use the AWS console . You can find it in the details of the wireguard EC2 instance, under IPv4 Public IP.

Login to Wireguard server via ssh. Remember that key authentication is enabled with the key pair already configured in AWS.

ssh ubuntu@

Now we install Wireguard on the server

sudo add-apt-repository -y ppa:wireguard/wireguard
sudo apt update
sudo apt install -y wireguard

Wireguard uses private and public keys for encryption.

We start creating a folder to store the keys

sudo mkdir /etc/wireguard/keys
sudo su -
cd /etc/wireguard/keys
umask 077

And then create the keys

wg genkey | tee privatekey | wg pubkey > publickey

The private key of the server must not be shared

Now we create the server config

nano /etc/wireguard/wg0.conf

With this content

PrivateKey = <server private key>
Address =
ListenPort = 51820

We are ready to start the wg0 interface

wg-quick up wg0
sudo systemctl enable wg-quick@wg0

Note the port 51820 of the server, it must match the port defined in the Security Group in our VPC

Now we can install Wireguard on the client.

After the client installation, collect its public key. We need the key to add it to the server using this command

sudo wg set wg0 peer insertClientPublicKeyHere persistent-keepalive 25 allowed-ips

Now we can save the configuration of Wireguard so they’ll survive a reboot of the server. Remember to do that after any change of the configuration, including adding new clients

sudo su -
wg showconf wg0 > /etc/wireguard/wg0.conf
wg setconf wg0 /etc/wireguard/wg0.conf

The server will accept the connections from the client and will only accept traffic originated from the IP address

To configure the client we need two pieces of information

  • the server public IP and port
  • the server public key

The public key is on the server. This key can and must be shared to permit the connection of the clients.

We get the key executing

cat /etc/wireguard/keys/publickey

The public IP of the server is the one we used to access it via ssh. We can get it via CLI running


Then we configure the client with the public IP and public Key of the server as shown in the screenshot

Raw content for copy/paste

PrivateKey = 09876543212098765432109876543212098765432210
Address =

PublicKey = 09876543212098765432109876543212098765432210
AllowedIPs =,
Endpoint =

Now we can activate the tunnel on the client. We should be able to ping the wg0 interface of the server but not the client yet.

Enable IP routing on the server

To reach the client with IP we need to enable IP forwarding on the Wireguard server

sysctl -w net.ipv4.ip_forward=1

To permanently enable IP routing add to /etc/sysctl.conf this line

net.ipv4.ip_forward = 1

And run

sysctl -p

Notice that in our Terraform template we disabled source/destination check for the Wireguard EC2 instance to permit IP forwarding in the VPC

source_dest_check = false

Almost there

We can now reach the internal EC2 instances through the VPN active in the public EC2 instance running Wireguard.

Ping to confirm it works.

We can ssh the client

ssh ubuntu@

and run a packet capture on the client to see the ping going in and out

sudo tcpdump -i eth0 net and icmp

The original IP of the VPN client is maintained, this can be useful for logging and troubleshooting.

If you’re interested to measure the performances, iperf can be the right tool for that.

Terraform destroy

When we’re done with the lab we must rember to run

terraform destroy

to remove all the resources created in AWS and avoid unexpected billing.

Wrap up

At the end of this post I run the same Terraform configuration using Terraform Cloud , reading the files form a GitHub private repository, with the same results. That is useful if you prefer not to install Terraform or to try something different.

Once again I experienced how a GUI can be easy for quick tasks or to create small infrastructures. For versioning, an approval workflow and anything that contains more than a few objects, IaC is the best way to do it (so far).

Writing a Terraform declarative configuration file requires more thinking about the purpose of every object. Working on a GUI feels shallow, but maybe it’s just me.

The Wireguard configuration for this post was very simple but it’s awesome to have a working setup with just a few commands. Wireguard can be used to create AWS Transit VPC design, but it still doesn’t work with AWS VPN . I’d not be suprised see Wireguard integrated in AWS in the future.

That’s it! Hope you enjoyed this post.