I am currently trying to deploy the UPF on AWS but as it is, there is no layer2 network I need to know how the traffic flows end to end so as to be able to know how to configure the network interfaces.
Though I was able to get the PODs running:
[ec2-user@ip-10-0-0-53 helm-chart]$ kubectl -n 5gcore get po -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
upf-c-7d95c9f87b-fjkjl 1/1 Running 0 50m 10.0.2.134 ip-10-0-2-130.eu-west-1.compute.internal <none> <none>
upf-lb-7f6f4957c7-g8vf2 1/1 Running 0 52m 10.0.2.11 ip-10-0-2-130.eu-west-1.compute.internal <none> <none>
upf-u-db4788888-6d7pz 1/1 Running 0 49m 10.0.2.157 ip-10-0-2-130.eu-west-1.compute.internal <none> <none>
But the PFCP return traffic from SMU seems not to leave the worker node..
Some information that might be helpful:
Is there need to configure the N4 IP directly within the SMU container?
Which component takes the input PFCP from the SMF? Is it the SMU or LBU?
How do I solve the EXT connectivity on the LBU since there is no layer2, maybe a dpdk interface can be dedicated to the N3, N4, N6 and N9 instead of using just one interface?
Maybe vhost-user can be used as this will make it very flexible for both op-prem and public cloud scenarios? The userspace-cni can achieve this via ovs-dpdk or vpp.
Hi,
I am currently trying to deploy the UPF on AWS but as it is, there is no layer2 network I need to know how the traffic flows end to end so as to be able to know how to configure the network interfaces.
Though I was able to get the PODs running:
But the PFCP return traffic from SMU seems not to leave the worker node..
Some information that might be helpful: