OpenAirInterface 5G Core and RAN Network Function Deployment using Helm Charts¶

OAI 5G core network have different network functions which can be used invidiually or deployed all together in different combination on a production grade Kubernetes cluster like Openshift or a Vanilla Kubernetes cluster.

Helm Chart Deployment

Reading time: ~30 mins

Tutorial replication time: ~40mins

NOTE: This tutorial will move in Orchestration repository starting next release.

Pre-requisite¶

Minimum 4 CPU (Threaded), 16GBi RAM and 50G storage required by NFs
(Optional) Multus CNI if using multiple interfaces for NFs

Clone the git repository

git clone -b <Branch> https://gitlab.eurecom.fr/oai/cn5g/oai-cn5g-fed

1. Description¶

The helm charts can be used on any production grade kubernetes cluster or even vanilla kubernetes. We have also tested on a single node 4 CPU and 16 GB ram minikube cluster with docker virtualization environment. In our testing environment we deploy these charts on our inhouse Openshift clusters the cluster information can be found below.

Software	Version
Openshift Client Version	4.10-4.16
Kubernetes Version	Kubernetes Version: v1.27.4
helm	v3.11.2
Base images of Network functions	Ubuntu 22.04/UBI 9(RHEL 9)

Each NF has its independent helm-chart and apart from that there are parent helm-charts for below scenarios:

Minimalist deployment: MYSQL (Subscriber Database), AMF, SMF, UPF, NRF
Basic deployment: MYSQL (Subscriber Database), UDR, UDM, AUSF, AMF, SMF, UPF, NRF, LMF
Advance deployment: MYSQL (Subscriber Database), NSSF, UDR, UDM, AUSF, AMF, SMF, UPF, NRF

End to End testing scenario: 1. Case 1: Monolithic RAN: OAI-gNB (RFSimulated), OAI-NR-UE (RFSimulated) 2. Case 2: F1 Split: OAI-CU, OAI-DU (RFSimulated), OAI-NR-UE (RFSimulated) 3. Case 3: E1-F1 Split: OAI-CU-CP, OAI-CU-UP, OAI-DU (RFSimulated), OAI-NR-UE (RFSimulated)

To know more about RFSimulator

Before you move forward lets create a namespace to host all the resources

kubectl apply -f - <<EOF
apiVersion: v1
kind: Namespace
metadata:
  name: oai-tutorial
  labels:
    pod-security.kubernetes.io/warn: "privileged"
    pod-security.kubernetes.io/audit: "privileged"
    pod-security.kubernetes.io/enforce: "privileged"
EOF

For Openshift you need to create a new project instead

oc new-project oai-tutorial

And later add the labels to allow creating privileged pods.

2. Fetching Network Function Images¶

Ubuntu base images can be pulled from docker-hub. In case you want to do some changes in the code then you should build your own images. If you will use Ubuntu images then skip this part and in section 3.1 there is a detailed procedure. In case of RHEL based worker node you can build your own UBI images, to download packages from RHEL repository you need a developer or enterprise account.

To learn how to build UBI 9.X images follow this tutorial
To learn how to build Ubuntu images follow this tutorial

3. Configuring Helm Charts¶

cd oai-cn5g-fed
ls charts/
oai-5g-core  oai-5g-ran
ls charts/oai-5g-core/
mysql  oai-5g-advance  oai-5g-basic  oai-5g-mini  oai-amf  oai-ausf  oai-lmf  oai-nrf  oai-nssf  oai-smf  oai-traffic-server  oai-udm  oai-udr  oai-upf
ls charts/oai-5g-ran/
oai-cu  oai-cu-cp  oai-cu-up  oai-du  oai-gnb  oai-nr-ue

All the OAI core network charts are present in oai-5g-core folder, there you can find charts of individual network functions and for the above described three different deployment settings.

Folder oai-5g-mini is for minimalist deployment
Folder oai-5g-basic is for basic deployment
Folder oai-5g-advance is for advance deployment

These charts are configured keeping in mind 5G service based architecture, if you want to deploy using reference based architecture then you need to make certain changes.

The structure of all these folders is similar,

oai-5g-basic/
├── Chart.yaml
├── config.yaml
├── README.md
├── templates
│   ├── configmap.yaml
│   ├── deployment.yaml
│   ├── _helpers.tpl
│   ├── multus.yaml (Only in AMF, SMF and UPF)
│   ├── NOTES.txt
│   ├── rbac.yaml
│   ├── serviceaccount.yaml
│   └── service.yaml
└── values.yaml

1 directory, 5 files

Starting version 2.0.0 of OAI 5G Core network functions their configuration will be in config.yaml and all infrastructure related information including image definition will be in values.yaml.

Helm chart of every network function looks similar and has the below structure. Only the chart of mysql database and NRF is different.

Network_function/
├── Chart.yaml
├── config.yaml
├── README.md
├── templates
│   ├── configmap.yaml
│   ├── deployment.yaml
│   ├── _helpers.tpl
│   ├── multus.yaml
│   ├── NOTES.txt
│   ├── rbac.yaml
│   ├── serviceaccount.yaml
│   └── service.yaml
└── values.yaml

1 directory, 12 files

All the configurable parameters for a particular commit/release are mentioned in the config.yaml file. These parameters will keep on changing in the future depending on the nature of development and features.

NOTE: If there is a need to edit a specific configuration parameter that is not configurable via the helm-chart config.yaml file then it has to be changed at the time of building images.

NOTE: Any changes done in the parent chart (Mini, basic, advance, case1, case2 or case3 scenario helm charts) will overwrite the sub charts.

Core Network functions discovers each-other using NRF and instead of using the ip-address of network functions we rely on using their FQDN, Kubernetes service concept. To communicate with each other whether we deploy them in reference point architecture or service based architecture.

For example: AMF registers with NRF using NRF FQDN (oai-nrf.oai-tutorial.svc.cluster.local). This way we can get rid of any static ip-address configuration.

3.1.1 Configure Multiple Interfaces¶

Here the network functions will use different virtual Ethernet interfaces to bind their different logical interface. Example AMF communicates with gNB/CU/CU-CP using N2 and with SMF and NRF using Namf, the Service Base Interface (SBI).
This type of configuration is also used when gNB/CU/CU-CP is outside of the cluster or UPF is outside of the cluster.
To make the above separation we are using multus to provide multiple Ethernet interfaces to network functions which have multiple communication interfaces.
Only AMF, SMF, UPF and RAN network functions have the possibility to use multus. Other network functions can also use multus but then it needs to be configured.
To configure multus for AMF, SMF, UPF or RAN network functions in values.yaml of each network function edit the multus section.

## Example from oai-amf/values.yaml
multus:
  ## If you don't want to add a default route in your pod then leave this field empty
  defaultGateway: "172.21.7.254"
  n2Interface:
    create: false
    ipAdd: "172.21.6.94"
    netmask: "22"
    ## If you do not have a gateway leave the field empty
    gateway:
    ## If you do not want to add any routes in your pod then leave this field empty
    routes: [{'dst': '10.8.0.0/24','gw': '172.21.7.254'}]
    name: 'n2'
    hostInterface: "bond0" # Interface of the host machine on which this pod will be scheduled

3.1.2 Use Single Interface¶

No need to configure multus for any network function. For different communication interfaces network functions will use same Ethernet interface. Example AMF will use eth0 interface to communicate with gNB, SMF and NRF.
In values.yaml of AMF, SMF and UPF in multus section do multus.create false like below,

## Example from oai-amf/values.yaml
multus:
  n2Interface:
    create: false

3.1.3 Capturing Packets (Optional)¶

Every network function has an extra TCP dump container to take the TCP dump. But by default this container is not used. If enabled it will capture dumps on all interfaces and will store inside the container locally or in a persistent volume if enabled.

To enable the persistent volume in the values.yaml of every network function make the below change,

## amf
start:
  amf: true #If false the network function container will run in sleep mode for manually testing
  tcpdump: true

includeTcpDumpContainer: true #If true it will add a tcpdump container inside network function pod for debugging

#To store PCAP of NF in a sharedVolume so it can be easily fetched (PVC is created with NRF charts so make sure in NRF it is true)
persistent:
  sharedvolume: true

3.2 Network function Images¶

To pull images from docker-hub, it will be good to configure an image pull secrete because of the anonymous pull limit of 100 in docker-hub. If you provide a personal/company account the limit will be different.

#kubernetes
kubectl create secret docker-registry regcred --docker-server=https://index.docker.io/v1/ --docker-username=<your-name> --docker-password=<your-pword> --docker-email=<your-email>
#openshift 
oc create secret docker-registry regcred --docker-server=https://index.docker.io/v1/ --docker-username=<your-name> --docker-password=<your-pword> --docker-email=<your-email>

There are more ways to make docker secrete you can follow this link. After this mention this in the values.yaml of each network function

## good to use when pulling images from docker-hub mention
imagePullSecrets:
  - name: "regcred"

When pulling images from docker hub you have several choices either to use images with develop tag (based on latest develop branch sometimes might not be stable), latest (built from current master branch) and release tags.

3.3 Configuring Helm Chart Parameters¶

In the config.yaml of oai-5g-basic helm charts you will see the configurable parameters for all the network functions check, the PLMN, DNN and subscriber information in mysql database

For basic and advance deployment check the database oai_db-basic.sql

A new subscriber entry can be added directly in the sql file or it can be added once the core network is already deployed.

To add the entry before deploying the core network, make sure you have all the required subscriber information IMSI(ueid/supi), Key(encPermanentKey), OPC(encOpcKey), PLMN, NSSAI(SST, SD), DNN

 vim/vi/nano charts/oai-5g-core/mysql/initialization/oai_db-basic.sql
# Add a new or edit existing entries after AuthenticationSubscription table
INSERT INTO `AuthenticationSubscription` (`ueid`, `authenticationMethod`, `encPermanentKey`, `protectionParameterId`, `sequenceNumber`, `authenticationManagementField`, `algorithmId`, `encOpcKey`, `encTopcKey`, `vectorGenerationInHss`, `n5gcAuthMethod`, `rgAuthenticationInd`, `supi`) VALUES
('208990100001124', '5G_AKA', 'fec86ba6eb707ed08905757b1bb44b8f', 'fec86ba6eb707ed08905757b1bb44b8f', '{\"sqn\": \"000000000020\", \"sqnScheme\": \"NON_TIME_BASED\", \"lastIndexes\": {\"ausf\": 0}}', '8000', 'milenage', 'c42449363bbad02b66d16bc975d77cc1', NULL, NULL, NULL, NULL, '208990100001124');
# Add the PDN/DNN information after SessionManagementSubscriptionData table
# To assign a static ip-address use the below entry
INSERT INTO `SessionManagementSubscriptionData` (`ueid`, `servingPlmnid`, `singleNssai`, `dnnConfigurations`) VALUES 
('208990100001124', '20899', '{\"sst\": 1, \"sd\": \"10203\"}','{\"oai\":{\"pduSessionTypes\":{ \"defaultSessionType\": \"IPV4\"},\"sscModes\": {\"defaultSscMode\": \"SSC_MODE_1\"},\"5gQosProfile\": {\"5qi\": 6,\"arp\":{\"priorityLevel\": 1,\"preemptCap\": \"NOT_PREEMPT\",\"preemptVuln\":\"NOT_PREEMPTABLE\"},\"priorityLevel\":1},\"sessionAmbr\":{\"uplink\":\"100Mbps\", \"downlink\":\"100Mbps\"},\"staticIpAddress\":[{\"ipv4Addr\": \"12.1.1.85\"}]}}');
INSERT INTO `SessionManagementSubscriptionData` (`ueid`, `servingPlmnid`, `singleNssai`, `dnnConfigurations`) VALUES 
('208990100001125', '20899', '{\"sst\": 1, \"sd\": \"10203\"}','{\"oai\":{\"pduSessionTypes\":{ \"defaultSessionType\": \"IPV4\"},\"sscModes\": {\"defaultSscMode\": \"SSC_MODE_1\"},\"5gQosProfile\": {\"5qi\": 6,\"arp\":{\"priorityLevel\": 1,\"preemptCap\": \"NOT_PREEMPT\",\"preemptVuln\":\"NOT_PREEMPTABLE\"},\"priorityLevel\":1},\"sessionAmbr\":{\"uplink\":\"100Mbps\", \"downlink\":\"100Mbps\"}}}');

In the config file smf.use_local_subscription_info should be no to use the user DNN subscription information from the database. Else it will be used from the configuration file.

Once the charts are configured perform helm dependency update inside the chart repository

cd charts/oai-5g-core
helm dependency update oai-5g-basic

NOTE: Whenever you will make any change in the network function helm-chart or mysql helm chart you need to perform a dependency update to inform parent chart about the sub-charts update.

4. Deploying Helm Charts¶

Once the configuration is finished the charts can be deployed with a user who has the rights to

Create RBAC (Optional only if Openshift is used)
Run pod (only UPF needs that) with privileged and anyuid scc (optional only required if you have scc configure in your cluster)
Create multus binds (optional only if multus is used)
Create namespace/project

helm install basic oai-5g-basic/ -n oai-tutorial

The output is similar to:

NAME: basic
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None

This command can take around ~3 mins depending on your network speed and cluster configuration (computational capacity). You can use the wait command to see if the core network functions are running or not

kubectl wait -n oai-tutorial --for=condition=ready pod -l app.kubernetes.io/instance=basic --timeout=3m

4.1 How to check if the Core network is properly configured?¶

Check the logs smf and upf to see that the PFCP session is properly configured,

kubectl logs -l app.kubernetes.io/name=oai-smf -n oai-tutorial | grep 'handle_receive(16 bytes)' | wc -l
kubectl logs -l app.kubernetes.io/name=oai-upf -n oai-tutorial | grep 'handle_receive(16 bytes)' | wc -l

If the value is more than 1 for both then it will verify that smf and upf have successfully registered to nrf and there is a PFCP session.

5 Use Case 1: Testing with Monolithic RAN¶

The images which are used in the tutorial are already present in docker-hub like the other images of OAI 5g core network functions. The charts of all the network functions are pre-configured to work with OAI-gnB and OAI-NR-UE end to end installation.

5.1 Images OAI-gNB and OAI-NR-UE¶

For Ubuntu based worker nodes the images can be pulled directly from docker-hub. To build images manually follow this link. In case you have an Openshift cluster then follow this link

5.2 Configuring OAI-gNB RFSimulator and OAI-NR-UE¶

Very Important To access internet in NR-UE the N6 interface of UPF should be able access the internet.

GNB requires the ip-address or service name of AMF. In case in AMF multus is used and N1/N2 interface is bind to multus interface, then please provide AMF ip-address.

For this tutorial we are not using multus here for simplicity, generally there should be two interfaces of gNB(for N2 and N3).

## oai-gNB configuration from values.yaml
config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: "--rfsim --log_config.global_log_options level,nocolor,time"
  gnbName: "oai-gnb-rfsim"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"  #currently only 4 standard values are allowed 1,2,3,4
  usrp: rfsim   #allowed values rfsim, b2xx, n3xx or x3xx
  amfhost: "oai-amf"  # amf ip-address or service-name oai-amf-svc or 172.21.6.94
  n2IfName: "eth0"    # if multus.n2Interface.create is true then use n2
  n3IfName: "eth0"   #if multus.n3Interface.create is true then use n3 or you can only use 1 interface n2 or eth0

5.3 Deploy OAI-gNB RFSimulator¶

To deploy the oai-gnb in rf simulator mode follow the below steps

cd ../oai-5g-ran/
helm install gnb oai-gnb --namespace oai-tutorial

The output is similar to:

NAME: gnb
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB_POD_NAME=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-gnb,app.kubernetes.io/instance=gnb" -o jsonpath="{.items[0].metadata.name}")
  export GNB_eth0_IP=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-gnb,app.kubernetes.io/instance=gnb" -o jsonpath="{.items[*].status.podIP}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. Note: This helm chart of OAI-gNB is only tested in RF-simulator mode and is not tested with USRPs/RUs on Openshift/Kubernetes Cluster
4. In case you want to test these charts with USRP/RU then make sure your underlying kernel is realtime and CPU sleep states are off. 
   Also for good performance it is better to use MTU 9000 for Fronthaul interface. 
5. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF

Wait for the gNB to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-gnb --timeout=3m --namespace oai-tutorial

To check if gnB is connected, read the logs of amf and check N2 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-amf| awk '{print $1}') | grep 'Connected'

The output is similar to:

Defaulted container "amf" out of: amf, init (init)
[2023-12-12 09:47:54.854] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-gnb-rfsim           |               001, 01              | 
[2023-12-12 09:48:14.854] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-gnb-rfsim           |               001, 01              |

5.4 Configure OAI-NR-UE RFSimulator¶

config:
  timeZone: "Europe/Paris"
  rfSimServer: "oai-ran"    # ip-address of rfsim or service name oai-gnb or oai-du
  fullImsi: "001010000000100"       # make sure all the below entries are present in the subscriber database
  fullKey: "fec86ba6eb707ed08905757b1bb44b8f"
  opc: "C42449363BBAD02B66D16BC975D77CC1"
  dnn: "oai"
  sst: "1"                     # configure according to gnb and amf, smf and upf 
  sd: "16777215"
  usrp: "rfsim"            # allowed rfsim, b2xx, n3xx, x3xx
  useAdditionalOptions: "--rfsim -r 106 --numerology 1 -C 3619200000 --log_config.global_log_options level,nocolor,time"

5.5 Deploy OAI-NR-UE RFSimulator¶

To deploy the oai-nr-ue in rf simulator mode follow the below steps

helm install nrue oai-nr-ue/ --namespace oai-tutorial

The output is similar to:

NAME: nrue
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export NR_UE_POD_NAME=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-nr-ue,app.kubernetes.io/instance=nrue" -o jsonpath="{.items[0].metadata.name}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. Note: This helm chart of OAI-NR-UE is only tested in RF-simulator mode not tested with hardware on Openshift/Kubernetes Cluster
4. In case you want to test these charts with USRP then make sure your CPU sleep states are off```
</details>

Wait for the NR-UE to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-nr-ue --timeout=3m --namespace oai-tutorial

Now you are start reading the logs of amf, smf and other network function to understand the message flow. Once the pdu session establishment procedure is finished you will receive an ip-address. You can start performing some testing.

check if the UE received an ip-address

kubectl exec -it -n oai-tutorial -c nr-ue $(kubectl get pods | grep oai-nr-ue | awk '{print $1}') -- ifconfig oaitun_ue1 |grep -E '(^|\s)inet($|\s)' | awk {'print $2'}

The output is similar to:

12.1.1.100

5.6 Performing some traffic testing¶

Inside the nr-ue pod there is an extra tcdump container which can be use to perform traffic testing via iperf3 or

kubectl exec -it -n oai-tutorial -c nr-ue $(kubectl get pods | grep oai-nr-ue | awk '{print $1}') -- ping -I oaitun_ue1 -c4 google.fr
PING google.fr (216.58.213.67) from 12.1.1.100 oaitun_ue1: 56(84) bytes of data.
64 bytes from par21s18-in-f3.1e100.net (216.58.213.67): icmp_seq=1 ttl=117 time=27.0 ms
64 bytes from par21s18-in-f3.1e100.net (216.58.213.67): icmp_seq=2 ttl=117 time=22.3 ms
64 bytes from par21s18-in-f3.1e100.net (216.58.213.67): icmp_seq=3 ttl=117 time=24.1 ms
64 bytes from par21s18-in-f3.1e100.net (216.58.213.67): icmp_seq=4 ttl=117 time=22.7 ms

--- google.fr ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3003ms
rtt min/avg/max/mdev = 22.375/24.072/27.031/1.833 ms

## incase above doesn't work try with 8.8.8.8 instead of dns. If that works then probably you have't configure dns properly in SMF.

NOTE: You can also deploy the Core and RAN network functions directly via the parent helm-chart

helm dependency update charts/e2e_scenarios/case1
helm install case1 charts/e2e_scenarios/case1 -n oai-tutorial

6 Use Case 2: Testing with F1 Split RAN¶

The images used for OAI-CU and OAI-DU are the same images used for OAI-gNB.

6.1 Configuration of OAI-CU and OAI-DU¶

Very Important To access internet in NR-UE the N6 interface of UPF should be able access the internet.

OAI-CU requires the ip-address or service name of AMF. In case in AMF multus is used and N1/N2 interface is bind to multus interface, then please provide AMF ip-address.

For this tutorial we are not using multus here for simplicity, generally there should be three interfaces of CU(for F1,N2, and N3).

## oai-cu configuration from values.yaml
config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: "--log_config.global_log_options level,nocolor,time"
  # If mounting the configuration file then below parameters are not used
  cuName: "oai-cu"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"     
  amfhost: "oai-amf"  # amf ip-address or service-name oai-amf-svc or 172.21.6.94
  n2IfName: "eth0"    # if multus.n2Interface.create is true then use n2
  n3IfName: "eth0"   #if multus.n3Interface.create is true then use n3 or you can only use 1 interface n2 or eth0 
  f1IfName: "eth0"   #if multus.f1Interface.create is true then use multus.f1Interface.Ipadd
  f1cuPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different
  f1duPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different

## oai-du configuration from values.yaml
config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: "--rfsim --log_config.global_log_options level,nocolor,time"
  duName: "oai-du-rfsim"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"  #currently only 4 standard values are allowed 1,2,3,4
  usrp: rfsim   #allowed values rfsim, b2xx, n3xx or x3xx
  f1IfName: "eth0"   #if multus.f1Interface.create is true then use f1
  cuHost: "oai-cu" ## Ip-address or hostname
  f1cuPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different
  f1duPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different

6.2 Deploy OAI-CU and OAI-DU (RFSimulator)¶

helm install cu oai-cu --namespace oai-tutorial

The output is similar to:

NAME: cu
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB__CU_POD_NAME=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-cu,app.kubernetes.io/instance=cu" -o jsonpath="{.items[0].metadata.name}")
  export GNB_CU_eth0_IP=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-cu,app.kubernetes.io/instance=cu" -o jsonpath="{.items[*].status.podIP}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF

Wait for the oai-cu to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-cu --namespace oai-tutorial --timeout=3m

Deploy oai-du

helm install du oai-du --namespace oai-tutorial

The output is similar to:

NAME: du
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB_DU_POD_NAME=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-du,app.kubernetes.io/instance=du" -o jsonpath="{.items[0].metadata.name}")
  export GNB_DU_eth0_IP=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-du,app.kubernetes.io/instance=du" -o jsonpath="{.items[*].status.podIP}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. Note: This helm chart of oai-du is only tested in RF-simulator mode and is not tested with USRPs/RUs on Openshift/Kubernetes Cluster
4. In case you want to test these charts with USRP/RU then make sure your underlying kernel is realtime and CPU sleep states are off
5. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF

Wait for the oai-du to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-du --namespace oai-tutorials --timeout=3m

To check if oai-du is connected to oai-cu, read the logs of cu and check F1 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-cu| awk '{print $1}') | grep 'Received F1 Setup Request'

The output is similar to:

838801.974035 [NR_RRC] I Received F1 Setup Request from gNB_DU 3584 (oai-du-rfsim) on assoc_id 189

To check if oai-cu is connected to amf, read the logs of amf and check N2 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-amf| awk '{print $1}') | grep 'Connected'

The output is similar to:

Defaulted container "amf" out of: amf, init (init)
[2023-12-12 10:49:54.879] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:14.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:34.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:54.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:51:14.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              |

After this follow the same procedure to start oai-nr-ue and ping to see if the UE is connected.

NOTE: You can also deploy the Core and RAN network functions directly via the parent helm-chart

helm dependency update charts/e2e_scenarios/case2
helm install case2 charts/e2e_scenarios/case2 -n oai-tutorial

7 Use Case 3: Testing with E1 and F1 Split RAN¶

The images used for OAI-CU-CP and OAI-DU are the same images used for OAI-gNB. OAI-CU-UP uses a different image which is present in docker-hub.

6.1 Configuration of OAI-CU and OAI-DU¶

Very Important To access internet in NR-UE the N6 interface of UPF should be able access the internet.

OAI-CU-CP requires the ip-address or service name of AMF. In case in AMF multus is used and N1/N2 interface is bind to multus interface, then please provide AMF ip-address.

For this tutorial we are not using multus here for simplicity, generally there should be three interfaces of CU-CP(for F1,N2 and E1).

config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: "--log_config.global_log_options level,nocolor,time"
  cucpName: "oai-cu-cp"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"  #currently only 4 standard values are allowed 1,2,3,4
  amfhost: "oai-amf"  # amf ip-address or service-name oai-amf-svc or 172.21.6.94
  n2IfName: "eth0"    # if multus.n2Interface.create is true then use n2
  f1IfName: "eth0"   #if multus.f1Interface.create is true then use f1
  e1IfName: "eth0"   #if multus.f1Interface.create is true then use e1
  f1cuPort: "2153"   #2153 if using same interface for f1 and n2 else standard port 2152 should be use if f1 and n3 interface are different
  f1duPort: "2153"   #2153 if using same interface for f1 and n2 else standard port 2152 should be use if f1 and n3 interface are different

## oai-cu-up configuration from values.yaml
config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: ""
  cuupName: "oai-cuup"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"     #currently only 4 standard values are allowed 1,2,3,4
  cuCpHost: "oai-cu" # 
  n3IfName: "eth0"   #if multus.n3Interface.create is true then use n3 or you can only use 1 interface n3 or eth0 
  f1IfName: "eth0"   #if multus.f1uInterface.create is true then use f1 or you can only use 1 interface n3 or eth0 
  e1IfName: "eth0"   #if multus.e1Interface.create is true then use e1 or you can only use 1 interface n3 or eth0 
  f1cuPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different
  f1duPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different

## oai-du configuration from values.yaml
config:
  timeZone: "Europe/Paris"
  useAdditionalOptions: "--rfsim --log_config.global_log_options level,nocolor,time"
  duName: "oai-du-rfsim"
  mcc: "001"   # check the information with AMF, SMF, UPF
  mnc: "01"    # check the information with AMF, SMF, UPF
  tac: "1"     # check the information with AMF
  sst: "1"  #currently only 4 standard values are allowed 1,2,3,4
  usrp: rfsim   #allowed values rfsim, b2xx, n3xx or x3xx
  f1IfName: "eth0"   #if multus.f1Interface.create is true then use f1
  cuHost: "oai-cu" ## Ip-address or hostname
  f1cuPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different
  f1duPort: "2153"   #2153 if using same interface for f1 and n3 else standard port 2152 should be use if f1 and n3 interface are different

7.1 Deploy OAI-CU-CP, OAI-CU-UP and OAI-DU¶

helm install cucp oai-cu-cp --namespace oai-tutorial

The output is similar to:

NAME: cucp
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB__CU_CP_POD_NAME=$(kubectl get pods --namespace default -l "app.kubernetes.io/name=oai-cu-cp,app.kubernetes.io/instance=cucp" -o jsonpath="{.items[0].metadata.name}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF

Wait for the oai-cu-cp to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-cu-cp --timeout=3m

Deploy oai-cu-up

helm install cuup oai-cu-up --namespace oai-tutorial

The output is similar to:

NAME: cuup
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB__CU_POD_NAME=$(kubectl get pods --namespace default -l "app.kubernetes.io/name=oai-cu-up,app.kubernetes.io/instance=cuup" -o jsonpath="{.items[0].metadata.name}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF
4. For good performance make sure your underlying kernel is realtime and CPU sleep states are off

Wait for the oai-cu-cp to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-cu-up --namespace oai-tutorial --timeout=3m

To check if oai-cu-up is connected to oai-cu-up, read the logs of cu-cp and check E1 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-cu-cp| awk '{print $1}') | grep 'Accepting new CU-UP ID'

The output is similar to:

840052.554668 [RRC] I Accepting new CU-UP ID 3584 name oai-cuup (assoc_id 193)

Deploy oai-du

helm install du oai-du --namespace oai-tutorial

The output is similar to:

NAME: du
NAMESPACE: oai-tutorial
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Get the application name by running these commands:
  export GNB_DU_POD_NAME=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-du,app.kubernetes.io/instance=du" -o jsonpath="{.items[0].metadata.name}")
  export GNB_DU_eth0_IP=$(kubectl get pods --namespace oai-tutorial -l "app.kubernetes.io/name=oai-du,app.kubernetes.io/instance=du" -o jsonpath="{.items[*].status.podIP}")
2. Dockerhub images of OpenAirInterface requires avx2 capabilities in the cpu and they are built for x86 architecture, tested on UBUNTU OS only.
3. Note: This helm chart of oai-du is only tested in RF-simulator mode and is not tested with USRPs/RUs on Openshift/Kubernetes Cluster
4. In case you want to test these charts with USRP/RU then make sure your underlying kernel is realtime and CPU sleep states are off
5. If you want to configure for a particular band then copy the configuration file in templates/configmap.yaml from here https://gitlab.eurecom.fr/oai/openairinterface5g/-/tree/develop/targets/PROJECTS/GENERIC-NR-5GC/CONF

Wait for the oai-du to start

kubectl wait --for=condition=ready pod -l app.kubernetes.io/name=oai-du --timeout=3m --namespace oai-tutorial

To check if oai-du is connected to oai-cu, read the logs of cu and check F1 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-cu-cp| awk '{print $1}') | grep 'Received F1 Setup Request'

The output is similar to:

838801.974035 [NR_RRC] I Received F1 Setup Request from gNB_DU 3584 (oai-du-rfsim) on assoc_id 189

To check if oai-cu-cp is connected to amf, read the logs of amf and check N2 setup procedure was correct or not,

kubectl logs --namespace oai-tutorial $(kubectl get pods --namespace oai-tutorial | grep oai-amf| awk '{print $1}') | grep 'Connected'

The output is similar to:

Defaulted container "amf" out of: amf, init (init)
[2023-12-12 10:49:54.879] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:14.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:34.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:50:54.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              | 
[2023-12-12 10:51:14.880] [amf_app] [info] |      1      |    Connected     |        0xe000        |       oai-cu           |               001, 01              |

After this follow the same procedure to start oai-nr-ue and ping to see if the UE is connected.

NOTE: You can also deploy the Core and RAN network functions directly via the parent helm-chart

helm dependency update charts/e2e_scenarios/case3
helm install case3 charts/e2e_scenarios/case3 -n oai-tutorial

8 Uninstall the helm charts¶

You can remove them one by one or you can use this command

helm uninstall -n oai-tutorial $(helm list -aq -n oai-tutorial)

9. Extra¶

9.1 How to capture pcaps inside a network function?¶

We have specially provided a separate container to capture pcap for each network function you can get inside this container and use tcpdump command

kubectl exec -it -c tcpdump $AMF_POD_NAME -- /bin/sh
tcpdump -i any -n

9.2 Resource Consumption of Network Functions¶

Below resource consumption is observered using Kubernetes metrics server while NR-UE was pinging to google.fr

Use case 1

case1-mysql-58f56f8f9b-ns5dj          15m          449Mi           
oai-amf-7c4ccf64f9-dgrgb              33m          5Mi             
oai-ausf-6fb7bb6b97-p9q9t             66m          3Mi             
oai-gnb-6d67985d99-zw5dx              1846m        1077Mi          
oai-nr-ue-697d6d8984-b5d2z            1378m        295Mi           
oai-nrf-9dffd9fc7-5k7nm               71m          3Mi             
oai-smf-6787fc5b9-qr82p               30m          5Mi             
oai-traffic-server-55db59f75d-kmzmw   5m           2Mi             
oai-udm-58c5879c5b-rz7jk              69m          3Mi             
oai-udr-98c5f7dd7-rkldg               65m          4Mi             
oai-upf-57c9b544ff-pst8j              25m          4Mi

Use case 2

kubectl get pods 
case2-mysql-546985bd6d-whg7x          15m          449Mi           
oai-amf-7775d6f48d-dzp7x              31m          5Mi             
oai-ausf-7f7fc9b766-llvr4             63m          3Mi             
oai-cu-67558c5bd5-9sdld               1m           11Mi            
oai-du-67d7f5766c-wdpn8               1827m        1074Mi          
oai-nr-ue-6dfb85f6df-mq88v            1355m        296Mi           
oai-nrf-7d6d4b4648-qzpf8              73m          3Mi             
oai-smf-d7bf796d6-z7r9g               28m          5Mi             
oai-traffic-server-55db59f75d-kfgpf   7m           2Mi             
oai-udm-86b4bdf7c7-bdcbc              66m          3Mi             
oai-udr-589758546b-7k8rq              69m          4Mi             
oai-upf-5594b9cc84-smjbq              27m          4Mi

Use case 3

kubectl top pods
NAME                                  CPU(cores)   MEMORY(bytes)   
case3-mysql-66db7fdcc6-q5rml          17m          450Mi           
oai-amf-574569c967-6dqnl              33m          5Mi             
oai-ausf-69b6444bff-qqvdr             70m          3Mi             
oai-cu-cp-8598c595-q42kh              1m           120Mi           
oai-cu-up-7988bb7f7c-pfnb4            1m           3Mi             
oai-du-56576c49d8-zdj5n               1827m        1074Mi          
oai-nr-ue-5fcb5cf474-x289m            1375m        300Mi           
oai-nrf-97b95548c-7tfkd               77m          3Mi             
oai-smf-d59bcb6c9-zkc89               26m          5Mi             
oai-traffic-server-55db59f75d-zjnnq   5m           2Mi             
oai-udm-ddf6dbc46-zhtn2               72m          3Mi             
oai-udr-67cf97b677-nk8gg              66m          4Mi             
oai-upf-5557c66469-79bqr              26m          4Mi

OpenAirInterface 5G Core and RAN Network Function Deployment using Helm Charts¶

Pre-requisite¶

1. Description¶

2. Fetching Network Function Images¶

3. Configuring Helm Charts¶

3.1 Networking related information¶

3.1.1 Configure Multiple Interfaces¶

3.1.2 Use Single Interface¶

3.1.3 Capturing Packets (Optional)¶

3.2 Network function Images¶

3.3 Configuring Helm Chart Parameters¶

4. Deploying Helm Charts¶

4.1 How to check if the Core network is properly configured?¶

5 Use Case 1: Testing with Monolithic RAN¶

5.1 Images OAI-gNB and OAI-NR-UE¶

5.2 Configuring OAI-gNB RFSimulator and OAI-NR-UE¶

5.3 Deploy OAI-gNB RFSimulator¶

5.4 Configure OAI-NR-UE RFSimulator¶

5.5 Deploy OAI-NR-UE RFSimulator¶

5.6 Performing some traffic testing¶

6 Use Case 2: Testing with F1 Split RAN¶

6.1 Configuration of OAI-CU and OAI-DU¶

6.2 Deploy OAI-CU and OAI-DU (RFSimulator)¶

7 Use Case 3: Testing with E1 and F1 Split RAN¶

6.1 Configuration of OAI-CU and OAI-DU¶

7.1 Deploy OAI-CU-CP, OAI-CU-UP and OAI-DU¶

8 Uninstall the helm charts¶

9. Extra¶

9.1 How to capture pcaps inside a network function?¶

9.2 Resource Consumption of Network Functions¶