Intro

In my previous [post]({{< ref "post/12-opnsense-virtualization-highly-available" >}}), I've set up a PoC to validate the possibility to create a cluster of 2 OPNsense VMs in Proxmox VE and make the firewall highly available.

Now I'm in the preparation to make it real in my homelab. So this time with a real setup, my future OPNsense cluster configuration.

Initially I was thinking of just restoring my current configuration on a OPNsense instance freshly installed. But then I realized that I didn't really documented how I put the pieces together the first time. This is the perfect moment to put things right.

In this post, I will show you how I configure OPNsense highly available, from a fresh installation in a couple of VMs, covering:

• Global settings
• Interfaces
• High Availability with CARP
• Firewall
• DHCP with Dnsmasq
• DNS with Unbound coupled with Dnsmasq
• VPN with WireGuard
• Reverse Proxy with Caddy
• And more...

Hopefully the next time, I will also cover the VM creation on Proxmox and how I'm preparing this migration from my physical OPNsense box to this highly available cluster in VMs. Let's dive in!

System

General

I start with the basics, in System > Settings > General:

• Hostname: cerbere-head1 (cerbere-head2 for the second one).
• Domain: mgmt.vezpi.com.
• Time zone: Europe/Paris.
• Language: English.
• Theme: opnsense-dark.
• Prefer IPv4 over IPv6: tick the box to prefer IPv4.

Users

Then, in System > Access > Users, I create a new user, I don't like sticking with the defaults root. I add this user in the admins group, while removing root from it.

Administration

In System > Settings > Administration, I change several things:

• TCP port: from 443 to 4443, to free port 443 for the reverse proxy coming next.
• HTTP Redirect: Disable to free port 80
• Alternate Hostnames: cerbere.vezpi.com which will be the URL to reach the firewall by the reverse proxy.
• Access log: enabled.
• Secure Shell Server: enabled.
• Authentication Method: permit password login (no root login).
• Sudo: No password. Once I click Save, I follow the link given to reach the WebGUI on port 4443.

Updates

Time for updates, in System > Firmware > Status, I click on Check for updates. An update is available, I close the banner, head to the bottom and click on Update. I'm warned that this update requires a reboot.

QEMU Guest Agent

Once updated and rebooted, I go to System > Firmware > Plugins, I tick the box to show community plugins. For now I only install the QEMU Guest Agent, os-qemu-guest-agent, to allow communication between the VM and the Proxmox host.

This requires a shutdown. On Proxmox, I enable the QEMU Guest Agent in the VM options:

Finally I restart the VM. Once started, from the Proxmox WebGUI, I can see the IPs of the VM which confirms the guest agent is working.

Interfaces

On both firewalls, I assign the remaining NICs to new interfaces adding a description. The VMs have 7 interfaces, I carefully compare the MAC addresses to not mix them up:

In the end, the interfaces configuration looks like this:

High Availability

Firewall Rule for pfSync

From here we can associate both instances to create a cluster. The last thing I need to do, is to allow the communication on the pfSync interface. By default, no communication is allowed on the new interfaces.

From Firewall > Rules > pfSync, I create a new rule on each firewall:

Configure HA

Next, I head to System > High Availability > Settings:

• Master (cerbere-head1):
  • Synchronize all states via: pfSync
  • Synchronize Peer IP: 192.168.44.2, the backup node IP
  • Synchronize Config: 192.168.44.2
  • Remote System Username: <username>
  • Remote System Password: <password>
  • Services: Select All
• Backup (cerbere-head2):
  • Synchronize all states via: pfSync
  • Synchronize Peer IP: 192.168.44.1, the master node IP
  • Synchronize Config: 192.168.44.1 ⚠️ Do not fill the XMLRPC Sync fields on the backup node, only to be filled on the master.

HA Status

In the section System > High Availability > Status, I can verify if the synchronization is working. On this page I can replicate any or all services from my master to my backup node:

Virtual IPs

Now that HA is configured, I can give my networks a virtual IP shared across my nodes. In Interfaces > Virtual IPs > Settings, I create one VIP for each of my networks using CARP (Common Address Redundancy Protocol). The target is to reuse the IP addresses used by my current OPNsense instance, but as it is still routing my network, I use different IPs for the configuration phase:

Firewall

Let's configure the core feature of OPNsense, the firewall. I don't want to go too crazy with the rules. I only need to configure the master, thanks to the replication.

Interface Groups

Basically I have 2 kinds of networks, those which I trust, and those which I don't. From this standpoint, I will create two zones.

Globally, on my untrusted networks, I will only allow access to the DNS and to the internet, no access towards other networks. On the other hand, my trusted networks would have the possibility to reach other VLANs.

To begin, in Firewall > Groups, I create 2 zones to regroup my interfaces:

• Trusted: Mgmt, User
• Untrusted: IoT, DMZ, Lab

Network Aliases

Next, in Firewall > Aliases, I create an alias InternalNetworks to regroup all my internal networks:

Firewall Rules

For all my networks, I want to allow DNS queries on the local DNS. In Firewall > Rules > Floating, let's create the first rule:

Next I want to allow connections towards the internet. At the same place I create a second rule:

Finally, I want to allow anything from my trusted networks. In Firewall > Rules > Trusted, I create the rule:

Great, with these 3 rules, I cover the basics. The remaining rules would be to allow specific equipment to reach out to something else. For example my home assistant instance want to connect to my TV, both are on different VLAN, hence I need a rule to allow it:

DHCP

For the DHCP, I choose Dnsmasq. In my current installation I use ISC DHCPv4, but as it is now deprecated, I prefer to replace it. Dnsmasq will also act as DNS, but only for my local zones.

Beware because it is not synchronizing leases in HA. To workaround this, both firewalls will serve DHCP at the same time, with slight different configuration to not overlap.

Dnsmasq General Configuration

In Services > Dnsmasq DNS & DHCP > General, I configure the master firewall as follow:

• Default
  • Enable: Yes
  • Interface: Mgmt, User, IoT, DMZ and Lab
• DNS
  • Listen port: 53053
• DNS Query Forwarding
  • Do not forward to system defined DNS servers: Enabled
• DHCP
  • DHCP FQDN: Enabled
  • DHCP local domain: Enabled
  • DHCP authoritative: Enabled
  • DHCP reply delay: 0
  • DHCP register firewall rules: Enabled
  • Disable HA sync: Enabled

On the backup node, I configure it the same, the only difference will be the DHCP reply delay which I set to 10. This will let the time to my master node to fulfill requests if it is alive.

DHCP Ranges

Next I configure the DHCP ranges. Both firewalls will have different ranges, the backup node will have smaller ones. On the master, they are configured as follow:

DHCP Options

Then I set some DHCP options for each domain: the router, the dns-server and the domain-name:

Hosts

Finally in in the Hosts tab, I define static DHCP mappings but also static IP not managed by the DHCP, to have them registered in the DNS:

DNS

For the DNS, I use Unbound. It is a validating, recursive, caching DNS resolver built into OPNsense, which can:

• Resolve queries from the root servers.
• Cache results for faster responses.
• Check domain authenticity with DNSSEC.
• Block domains based of blacklist.
• Add custom records.

For the local zones, I forward the requests to Dnsmasq, hence I will not registering DHCP leases in Unbound.

Unbound General Settings

Let's configure it, in Services > Unbound DNS > General:

DNS Blocklist

Then I configure the blocklist in Services > Unbound DNS > Blocklist. I enable it and select the [hagezi] Multi PRO mini list. Initially I was using AdGuard Home, but I want to give this blocklist feature a chance.

To maintain the service up to date, in System > Settings > Cron, I add my first job that runs every night at 2AM to Update Unbound DNSBLs.

Query Forwarding

Finally I configure query forwarding for my local domains to Dnsmasq. In Services > Unbound DNS > Query Forwarding, I add each of my local domains with their reverse lookup (PTR record):

VPN

When I'm not home, I still want to be able to reach my services and enjoy my DNS ad blocker. For that I'm setting up a VPN, with WireGuard. It's fast, secure and easy to set up.

WireGuard Instance Setup

In VPN > WireGuard > Instances, I create a new one:

• Enabled: Yes
• Name: Homelan
• Public/Private keys: Key-pair generated
• Listen port: 61337
• Tunnel address: 10.13.37.1/24
• Depend on (CARP): on lan (vhid 1)

Once configured, I enable WireGuard and apply the configuration.

Peer Setup

Next in the Peer generator tab, I fulfill the empty fields for my first device:

• Endpoint: vezpi.com
• Name: S25Ultra
• DNS Servers: 10.13.37.1

Before clicking Store and generate next, from my device, I configure the peer by capturing the QR code. Finally I can save that peer and start over for new ones.

Create VPN Interface

This step is not required, but ease the configuration management for firewall rules. On both firewall, in Interfaces > Assignments, I assign the wg0 (WireGuard - Homelan) interface and name it VPN.

Then in Interfaces > VPN, I enable this interface.

Finally, in Firewall > Groups, I add this interface in the Trusted group.

Firewall Rule

To allow connections from outside, I need to create a firewall rule on the WAN interface:

Reverse Proxy

The next feature I need is a reverse proxy, to redirect incoming HTTPS requests to reach my services, such as this blog. For that I use Caddy. This service is not installed by default, I need to add a plugin.

On both firewalls, In System > Firmware > Plugins, I tick the box to show community plugins and install os-caddy.

Caddy General Settings

I refresh the page and, on the master, in Services > Caddy > General Settings:

• Enable Caddy: Yes
• Enable Layer4 Proxy: Yes
• ACME: <email address>
• Auto HTTPS: On (default)

There are two types of redirections, the Reverse Proxy and the Layer4 Proxy. The first one is for HTTPS only, where Caddy will manage the SSL.

Reverse Proxy

In Services > Caddy > Reverse Proxy, I define the services directly managed by Caddy.

These services should not be exposed to everyone. In the Access tab, I create a list, called Internal, of allowed networks, including my LAN and VPN networks.

Then in the Domains tab, I add my domains. For example, this is here I define cerbere.vezpi.com, my URL to reach my OPNsense WebGUI:

• Enable: Yes
• Frontend
  • Protocol: https://
  • Domain: cerbere.vezpi.com
  • Port: leave empty
  • Certificate: Auto HTTPS
  • Description: OPNsense
• Access
  • Access List: Internal
  • HTTP Access Log: Enabled

Finally in the Handlers tab, I define to which upstream these domains are forwarded to. For cerbere.vezpi.com I define this:

• Enabled: Yes
• Frontend
  • Domain: https://cerbere.vezpi.com
  • Subdomain: None
• Handler
  • Path: any
• Access
  • Access List: None
• Directive
  • Directive: reverse_proxy
• Upstream
  • Protocol: https://
  • Upstream Domain: 127.0.0.1
  • Upstream Port: 4443
  • TLS Insecure Skip Verify: Enabled
  • Description: OPNSense

Layer4 Proxy

Most of my services are behind another reverse proxy on my network, Traefik. To let it manage normally its domains, I forward them using Layer4 Routes. It prevents Caddy to terminate SSL, the HTTPS stream is left intact.

In Services > Caddy > Layer4 Proxy, I create 3 routes.

The first one is for internet exposed services, like this blog or my Gitea instance:

• Enabled: Yes
• Sequence: 1
• Layer 4
  • Routing Type: listener_wrappers
• Layer 7
  • Matchers: TLS (SNI Client Hello)
  • Domain: blog.vezpi.com git.vezpi.com
  • Terminate SSL: No
• Upstream
  • Upstream Domain: 192.168.66.50
  • Upstream Port: 443
  • Proxy Protocol: v2
  • Description: External Traefik HTTPS dockerVM

The second one is for internal only services. It is configured pretty much the same but using access list:

• Sequence: 2
• Access
  • Remote IP: 192.168.13.0/24 192.168.88.0/24 10.13.37.0/24

The third one is for Traefik HTTP challenges for Let's Encrypt:

• Sequence: 3
• Layer 7
  • Matchers: HTTP (Host Header)
  • Domain: blog.vezpi.com git.vezpi.com etc.
• Upstream:
  • Upstream Port: 80
  • Proxy Protocol: Off (default)

Firewall Rules

Finally, I need to allow connection of these ports on the firewall, one rule for HTTPS and another for HTTP:

mDNS Repeater

The last service I want to setup in OPNsense is a mDNS repeater. This is useful for some devices to announce themselves on the network, when not on the same VLAN, such as my printer or my Chromecast. The mDNS repeater get the message from an interface to forward it to another one.

This service is also not installed by default, on both firewalls, In System > Firmware > Plugins, I tick the box to show community plugins and install os-mdns-repeater.

Then in Services > mDNS Repeater, the configuration is pretty straight forward:

• Enable: Yes
• Enable CARP Failover: Yes
• Listen Interfaces: IoT, User

CARP Failover Script

In my setup, I only have a single WAN IP address which is served by the DHCP of my ISP box. OPNsense does not provide natively a way to handle this scenario. To manage it, I implement the same trick I used in the [PoC]({{< ref "post/12-opnsense-virtualization-highly-available" >}}).

I copy the MAC of the net1 interface of cerbere-head1 and paste it to the same interface for cerbere-head2. Doing so, the DHCP lease for the WAN IP address can be shared among the nodes.

⚠️ Warning: Having two machines on the network with the same MAC can cause ARP conflicts and break connectivity. Only one VM should keep its interface active.

Under the hood, in OPNsense, a CARP event triggers some scripts. These are located in /usr/local/etc/rc.syshook.d/carp/. To manage WAN interface on each node, I implement this PHP script 10-wan on both nodes. Depending on their role (master or backup), this will enable or disable their WAN interface:

#!/usr/local/bin/php
<?php

require_once("config.inc");
require_once("interfaces.inc");
require_once("util.inc");
require_once("system.inc");

$subsystem = !empty($argv[1]) ? $argv[1] : '';
$type = !empty($argv[2]) ? $argv[2] : '';

if ($type != 'MASTER' && $type != 'BACKUP') {
    log_error("Carp '$type' event unknown from source '{$subsystem}'");
    exit(1);
}

if (!strstr($subsystem, '@')) {
    log_error("Carp '$type' event triggered from wrong source '{$subsystem}'");
    exit(1);
}

$ifkey = 'wan';

if ($type === "MASTER") {
    log_error("enable interface '$ifkey' due CARP event '$type'");
    $config['interfaces'][$ifkey]['enable'] = '1';
    write_config("enable interface '$ifkey' due CARP event '$type'", false);
    interface_configure(false, $ifkey, false, false);
} else {
    log_error("disable interface '$ifkey' due CARP event '$type'");
    unset($config['interfaces'][$ifkey]['enable']);
    write_config("disable interface '$ifkey' due CARP event '$type'", false);
    interface_configure(false, $ifkey, false, false);
}

Service Synchronization

The final step is to synchronize all the services between the master and the backup node in the cluster. First in System > High Availability > Status, I click the button to Synchronize and reconfigure all.

Then I want to make sure that future changes are synchronized if I omit to replicate them myself. In System > Settings > Cron, I add a new job that runs every night to HA update and reconfigure backup.

Conclusion