The way to restart only NF which was stopped

Q/A
1

Hello,
I was wondering if it is possible to resume the service by only restarting NF which was stopped.

According to Document of free5gc, it was mentioned about the boot order like blow.

NFs should be run in priority as below:
NRF > UDR > UDM > AUSF > NSSF > AMF > PCF > UPF > SMF > N3IWF

If any NF stops after booting, do I have to start from the beginning according to the boot order?

If anyone have any information about this topic, please let me know.
I appreciate for your help.

6

Hi @NK636,
For the problem about “Is there any way to resume the service only restarting the specific NF”, I have tested restarting SMF by killing original SMF process and run another new SMF process.
The procedure I tested is as following :
Run free5GC -> kill SMF -> UE registration is not working -> only restart a new SMF -> UE registration successfully
So, there is a way to resume the service only restarting the specific NF.

But about restarting NRF error, I consider the reason is that other NFs’ consumer need to register to NRF.
The new NRF doesn’t have management information of old NFs, so the old NFs can not access other registered NFs’ consumer

7

Hello @SymbolWu,

Thank you for your response.
I understood that we can resume the service only restarting the specific NF for SMF, and it would depends on NFs whether it is possible.

I might try the procedure you gave me about SMF restarting.

I appreciate for your help.

2

You don’t have to do it exactly in order, I can do it all at once.

3

Hello @xiaobinggan-zmh,

Thank you for your response.
I was concerned about the problem since I experienced NRF sudden stop, and restarting only NRF didn’t work.
AMF couldn’t communicate with NRF.

“Is there any way to resume the service only restarting the specific NF”, is the thing I wanted to know.
Do you have any idea about this?

I appreciate for your help.

4

Because I deploy with K8S and Docker, I specify the IP of NF in all configuration files as fixed. If they find each other by IP, they can start together without strict order.

5

Hello @xiaobinggan-zmh ,

Thank you for your reply.

So my understanding is, we don’t have to consider NF’s dependencies when using K8S or docker.
Is this correct?

I assume you are mentioning about “Service” functions on K8S/docker.
In that case I wonder if it is possible to resume even all pods are down in whatever service.
Do you have any idea?

Using K8s or dockers function sounded effective, so I thought I might try using it.

Thank you for your help.