* Detect current network and netmask Previously the network was guessed (assumed it was a /24 by replacing the last octet with a 0) and then later the netmask was hardcoded to 255.255.240.0 as a result the requested route is almost always incorrect. As most people supply their own or use the vpn tunnel as a default gateway it's probably rare to cause any issues. Signed-off-by: Harold Drost <baelish@bluecell.net> * Removing unnecessary debug messages. Signed-off-by: Harold Drost <baelish@bluecell.net> * Actually fixing the thing it was all about... :$ My testing worked because helm believed the `openvpn.conf` unchanged and therefore didn't replace the manually edited file which had the change in this commit. Tested using brand new deployment now works correctly and without routing errors. Signed-off-by: Harold Drost <baelish@bluecell.net>
Helm chart for OpenVPN
This chart will install an OpenVPN server inside a kubernetes cluster. New certificates are generated on install, and a script is provided to generate client keys as needed. The chart will automatically configure dns to use kube-dns and route all network traffic to kubernetes pods and services through the vpn. By connecting to this vpn a host is effectively inside a cluster's network.
Uses
The primary purpose of this chart was to make it easy to access kubernetes services during development. It could also be used for any service that only needs to be accessed through a vpn or as a standard vpn.
Usage
helm repo add stable http://storage.googleapis.com/kubernetes-charts
helm install stable/openvpn
Wait for the external load balancer IP to become available. Check service status via: kubectl get svc
Please be aware that certificate generation is variable and may take some time (minutes).
Check pod status, replacing $HELM_RELEASE with the name of your release, via:
POD_NAME=$(kubectl get pods -l "app=openvpn,release=$HELM_RELEASE" -o jsonpath='{.items[0].metadata.name}') \
&& kubectl logs "$POD_NAME" --follow
When all components of the openvpn chart have started use the following script to generate a client key:
#!/bin/bash
if [ $# -ne 3 ]
then
echo "Usage: $0 <CLIENT_KEY_NAME> <NAMESPACE> <HELM_RELEASE>"
exit
fi
KEY_NAME=$1
NAMESPACE=$2
HELM_RELEASE=$3
POD_NAME=$(kubectl get pods -n "$NAMESPACE" -l "app=openvpn,release=$HELM_RELEASE" -o jsonpath='{.items[0].metadata.name}')
SERVICE_NAME=$(kubectl get svc -n "$NAMESPACE" -l "app=openvpn,release=$HELM_RELEASE" -o jsonpath='{.items[0].metadata.name}')
SERVICE_IP=$(kubectl get svc -n "$NAMESPACE" "$SERVICE_NAME" -o go-template='{{range $k, $v := (index .status.loadBalancer.ingress 0)}}{{$v}}{{end}}')
kubectl -n "$NAMESPACE" exec -it "$POD_NAME" /etc/openvpn/setup/newClientCert.sh "$KEY_NAME" "$SERVICE_IP"
kubectl -n "$NAMESPACE" exec -it "$POD_NAME" cat "/etc/openvpn/certs/pki/$KEY_NAME.ovpn" > "$KEY_NAME.ovpn"
The entire list of helper scripts can be found on templates/config-openvpn.yaml
Be sure to change KEY_NAME if generating additional keys. Import the .ovpn file into your favorite openvpn tool like tunnelblick and verify connectivity.
Configuration
The following table lists the configurable parameters of the openvpn chart and their default values,
and can be overwritten via the helm --set flag.
| Parameter | Description | Default |
|---|---|---|
replicaCount |
amount of parallel openvpn replicas to be started | 1 |
updateStrategy |
update strategy for deployment | {} |
image.repository |
openvpn image repository |
jfelten/openvpn-docker |
image.tag |
openvpn image tag |
1.1.0 |
image.pullPolicy |
Image pull policy | IfNotPresent |
service.type |
k8s service type exposing ports, e.g. NodePort |
LoadBalancer |
service.externalPort |
TCP port reported when creating configuration files | 443 |
service.internalPort |
TCP port on which the service works | 443 |
service.nodePort |
NodePort value if service.type is NodePort |
nil (auto-assigned) |
service.externalIPs |
External IPs to listen on | [] |
resources.requests.cpu |
OpenVPN cpu request | 300m |
resources.requests.memory |
OpenVPN memory request | 128Mi |
resources.limits.cpu |
OpenVPN cpu limit | 300m |
resources.limits.memory |
OpenVPN memory limit | 128Mi |
persistence.enabled |
Use a PVC to persist configuration | true |
persistence.subPath |
Subdirectory of the volume to mount at | nil |
persistence.existingClaim |
Provide an existing PersistentVolumeClaim | nil |
persistence.storageClass |
Storage class of backing PVC | nil |
persistence.accessMode |
Use volume as ReadOnly or ReadWrite | ReadWriteOnce |
persistence.size |
Size of data volume | 2M |
podAnnotations |
Key-value pairs to add as pod annotations | {} |
openvpn.OVPN_NETWORK |
Network allocated for openvpn clients | 10.240.0.0 |
openvpn.OVPN_SUBNET |
Network subnet allocated for openvpn | 255.255.0.0 |
openvpn.OVPN_PROTO |
Protocol used by openvpn tcp or udp | tcp |
openvpn.OVPN_K8S_POD_NETWORK |
Kubernetes pod network (optional) | 10.0.0.0 |
openvpn.OVPN_K8S_POD_SUBNET |
Kubernetes pod network subnet (optional) | 255.0.0.0 |
openvpn.OVPN_K8S_SVC_NETWORK |
Kubernetes service network (optional) | nil |
openvpn.OVPN_K8S_SVC_SUBNET |
Kubernetes service network subnet (optional) | nil |
openvpn.dhcpOptionDomain |
Push a dhcp-option DOMAIN config |
true |
openvpn.conf |
Arbitrary lines appended to the end of the server configuration file | nil |
openvpn.redirectGateway |
Redirect all client traffic through VPN | true |
nodeSelector |
Node labels for pod assignment | {} |
This chart has been engineered to use kube-dns and route all network traffic to kubernetes pods and services,
to disable this behaviour set openvpn.OVPN_K8S_POD_NETWORK and openvpn.OVPN_K8S_POD_SUBNET to null.
If openvpn.OVPN_K8S_SVC_NETWORK and openvpn.OVPN_K8S_SVC_SUBNET are defined, an extra route for services subnet will be added.
Note: As configured the chart will create a route for a large 10.0.0.0/8 network that may cause issues if that is your local network. If so tweak this value to something more restrictive. This route is added, because GKE generates pods with IPs in this range.
Certificates
New certificates are generated with each deployment, if keystoreSecret is not defined. If persistence is enabled certificate data will be persisted across pod restarts. Otherwise new client certs will be needed after each deployment or pod restart.
Certificates can be passed in secret, which name is specified in openvpn.keystoreSecret value. Create secret as follows:
kubectl create secret generic openvpn-keystore-secret --from-file=./server.key --from-file=./ca.crt --from-file=./server.crt --from-file=./dh.pem
You can deploy temporary openvpn chart, create secret from generated certificates, and then re-deploy openvpn, providing the secret. Certificates can be found in openvpn pod in the following files:
/etc/openvpn/certs/pki/private/server.key
/etc/openvpn/certs/pki/ca.crt
/etc/openvpn/certs/pki/issued/server.crt
/etc/openvpn/certs/pki/dh.pem