This guideline shows how to deploy the free5gc on a Kubernetes cluster and then test it with UERANSIM.
- A Kubernetes cluster supporting SCTP
- Kubernetes worker nodes with kernel 5.0.0-23-generic and containing gtp5g kernel module (required for the Free5GC UPF element).
- Multus-CNI.
- Helm3.
- Kubectl (optional).
- A Persistent Volume (size 8Gi).
- A physical network interface on each Kubernetes node named eth0.
- A physical network interface on each Kubernetes node named eth1 to connect the UPF to the Data Network.
Note: If the names of network interfaces on your Kubernetes nodes are different from eth0
and eth1
, see Networks configuration.
Please refer to this section Networks configuration to make sure you'll not have a networking related issue.
First check that the Linux kernel version on Kubernetes worker nodes is 5.0.0-23-generic
or 5.4.x
.
uname -r
Then, on each worker node, install the gtp5g kernel module.
git clone -b v0.3.1 https://github.com/free5gc/gtp5g.git
cd gtp5g
make
sudo make install
If you don't have a Persistent Volume provisioner, you can use the following commands to create a namespace for the project and a Persistent Volume within this namespace that will be consumed by MongoDB by adapting it to your implementation (you have to replace worker1
by the name of the node and /home/vagrant/kubedata
by the right directory on this node in which you want to persist the MongoDB data).
kubectl create ns <namespace>
cat <<EOF | kubectl apply -f -
apiVersion: v1
kind: PersistentVolume
metadata:
name: example-local-pv9
labels:
project: free5gc
spec:
capacity:
storage: 8Gi
accessModes:
- ReadWriteOnce
persistentVolumeReclaimPolicy: Retain
local:
path: /home/vagrant/kubedata
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: kubernetes.io/hostname
operator: In
values:
- worker1
EOF
NOTE: you must create the folder on the right node before creating the Peristent Volume.
On the charts directory, run:
helm -n <namespace> install <free5GC-release-name> ./free5gc/
kubectl -n <namespace> get pods -l "project=free5gc"
The WEBUI can is exposed with a Kubernetes service with nodePort=30500
. So you can access it by using this url {replace-by-the-IP-of-one-of-your-cluster-nodes}:30500
.
For adding a new subscriber, please refer to the Free5GC documentation.
On the charts directory, run:
helm -n <namespace> install <UERANSIM-release-name> ./ueransim/
kubectl -n <namespace> get pods -l "app=ueransim"
Once the UERANSIM components created, you can access to the UE pod by running:
kubectl -n <namespace> exec -it <ue-pod-name> -- bash
Then, you can use the created TUN interface for more advanced testing. Please refer to the UERANSIM helm chart's README and check this link and the UERANSIM chart Readme for more details.
# Run this inside the container
ip address
...
5: uesimtun0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UNKNOWN group default qlen 500
link/none
inet 10.1.0.1/32 scope global uesimtun0
valid_lft forever preferred_lft forever
ping -I uesimtun0 www.google.com
traceroute -i uesimtun0 www.google.com
curl --interface uesimtun0 www.google.com
Check the readme of each Helm chart in the project to list configurable parameters.
According to the Free5GC documentation, you may sometimes need to drop the data stored in the MongoDB. To do so with our implementation, you need simply to empty the folder that was used in the Persistent Volume on the corresponding node.
sudo rm -rf {path-to-folder}/*
This may occur because of ipv4.ip_forward
being disabled on the UPF POD. In fact, this functionalty is needed by the UPF as it allows him to act as a router.
To check if it is enabled, run this command on the UPF POD. The result must be 1.
cat /proc/sys/net/ipv4/ip_forward
We remind you that some CNI plugins (e.g. Flannel) allow this functionality by default, while others (.e.g. Calico) require a special configuration.
Promiscuous mode must be enabled for master interfaces of MACVLAN interfaces. In fact, allows network interfaces to intercept packets even if the MAC destination address on these packets is different from the MAC address of this interface. This is necesary as MACVLAN interfaces we are using will have diffrent MAC addresses from MAC addresses of their master interfaces.
NOTE: how to enable Promiscuous highly depends on the technology or the tool used for the creation of VMs. Thus, you should check if it is enabled by default in your platform or how to enable it.