Integrate Kubernetes and pfSense
kubernetes-pfsense-controller (kpc) works hard to keep pfSense and Kubernetes in sync and harmony. The primary focus is to facilitate a first-class Kubernetes cluster by integrating and/or implementing features that generally do not come with bare-metal installation(s).
This is generally achieved using the standard Kubernetes API along with the xmlrpc API for pfSense. Speaking broadly
the Kubernetes API is
watched and appropriate updates are sent to pfSense (
config.xml) via xmlrpc calls along with
appropriate reload/restart/update/sync actions to apply changes.
Please note, this controller is not designed to run multiple instances simultaneously (ie: do NOT crank up the replicas).
Disclaimer: this is new software bound to have bugs. Please make a backup before using it as it may eat your configuration. Having said that, all known code paths appear to be solid and working without issue. If you find a bug, please report it!
The controller is comprised of several plugins that are enabled/disabled/configured via a Kubernetes ConfigMap. Details about each plugin follows below.
Services in Kubernetes. This is done via any
combination of Layer2 or BGP type configurations. Layer2 requires no integration with pfSense, however, if you want to
leverage the BGP implementation you need a BGP server along with neighbor configuration.
kpc dynamically updates
bgp neighbors for you in pfSense by continually monitoring cluster
The plugin assumes you've already installed openbgp and configured it as well as created a
group to use with MetalLB.
metallb: enabled: true nodeLabelSelector: nodeFieldSelector: bgp-implementation: openbgp options: openbgp: template: md5sigkey: md5sigpass: groupname: metallb row: - parameters: announce all parmvalue:
haproxy-declarative plugin allows you to declaratively create HAProxy frontend/backend definitions as
resources in the cluster. When declaring backends however, the pool of servers can/will be dynamically created/updated
based on cluster nodes. See declarative-example.yaml for an example.
haproxy-declarative: enabled: true
haproxy-ingress-proxy plugin allows you to mirror cluster ingress rules handled by an ingress controller to HAProxy
running on pfSense. If you run pfSense on the network edge with non-cluster services already running, you now can
dynamically inject new rules to route traffic into your cluster while simultaneously running non-cluster services.
To achieve this goal, new 'shared' HAProxy frontends are created and attached to an existing HAProxy frontend. Each
created frontend should also set an existing backend. Note that existing frontend(s)/backend(s) can be created manually
or using the
haproxy-declarative you can create a dynamic backend service (ie: your ingress controller) and
subsequently dynamic frontend services based off of cluster ingresses. This is generally helpful when you cannot or do
not for whatever reason create wildcard frontend(s) to handle incoming traffic in HAProxy on pfSense.
Optionally, on the ingress resources you can set the following annotations:
haproxy-ingress-proxy.pfsense.org/backend to respectively set the frontend and backend to override the defaults.
haproxy-ingress-proxy: enabled: true ingressLabelSelector: ingressFieldSelector: defaultFrontend: http-80 defaultBackend: traefik #allowedHostRegex: "/.*/"
kpc provides various options to manage DNS entries in pfSense based on cluster state. Note that these options can be
used in place of or in conjunction with external-dns to support
pfsense-dns-services watches for services of type
LoadBalancer that have the annotation
with the value of the desired hostname.
kpc will create the DNS entry in unbound/dnsmasq. Note that to actually get
an IP on these services you'll likely need
MetalLB deployed in the cluster (regardless of the
metallb plugin running
pfsense-dns-services: enabled: true serviceLabelSelector: serviceFieldSelector: #allowedHostRegex: "/.*/" dnsBackends: dnsmasq: enabled: true unbound: enabled: true
pfsense-dns-ingresses watches ingresses and automatically creates DNS entries in unbound/dnsmasq. This requires proper
support from the ingress controller to set IPs on the ingress resources.
pfsense-dns-ingresses: enabled: true ingressLabelSelector: ingressFieldSelector: #allowedHostRegex: "/.*/" dnsBackends: dnsmasq: enabled: true unbound: enabled: true
pfsense-dns-haproxy-ingress-proxy monitors the HAProxy rules created by the
haproxy-ingress-proxy plugin and creates
host aliases for each entry. To do so you create an arbitrary host in unbound/dnsmasq (something like
<frontend name>.k8s) and bind that host to the frontend through the config option
frontends.<frontend name>. Any
proxy rules created for that frontend will now automatically get added as aliases to the configured
sure the static
hostname created in your DNS service of choice points to the/an IP bound to the corresponding
pfsense-dns-haproxy-ingress-proxy: enabled: true #allowedHostRegex: "/.*/" dnsBackends: dnsmasq: enabled: true unbound: enabled: true frontends: http-80: hostname: http-80.k8s primary_frontend_name2: hostname: primary_frontend_name2.k8s
regex parameters are passed through php's
preg_match() method, you can test your syntax using that. Also note that
if you want to specify a regex ending (
$), you must escape it in yaml as 2
kpc stores it's stateful data in the cluster as a ConfigMap (kube-system.kubernetes-pfsense-controller-store by
default). You can review the data there to gain understanding into what the controller is managing.
You may need/want to bump up the
webConfigurator setting for
Max Processes to ensure enough simultaneous connections
can be established. Each
kpc instance will only require 1 process (ie: access to the API is serialized by
- base64 advanced fields (haproxy)
- taint haproxy config so it shows 'apply' button in interface?
- _index and id management
- ssl certs name/serial
- build docker images
- create manifests
- ensure pfsync items are pushed as appropriate
- perform config rollbacks when appropriate?
- validate configuration(s) to ensure proper schema
XML config structure (note that
ha_backends is actually frontends...it's badly named):
haproxy ha_backends item item ... ha_pools item ha_servers item item ... item ...