open5gs open5gcore fix use site DNS server for UE

[bjoern-r]

Change the 5g-core SMF configs to use a site configured DNS server instead of 8.8.8.8 when configuring the PDU session.

[bjoern-r]

TODO: changelog

[bjoern-r]

@alvarocurt could you maybe test the k8s changes?

[alvarocurt]

Hi Bjorn, Looks great, but i have a few regards:

• Why declare the DNS as a site variable? It depends on the intended use case of changing the UE DNS
  • If experimenter may want to play with the UE DNS value, it should be a public variable
  • If each site/platform wants to FORCE the UE DNS, it should be a site variable
• open5gs_k8s lacks some changes, but I can take care of them when I test it:
  • README is not updated as the other components.
  • public.yaml is not updated as the other components.
  • no need to define variable ue_dns_server to use it with the values template. Defaults can be ensured in jinja too.

BR!

[alvarocurt]

Requesting feedback on my latest comment before accepting.

[bjoern-r]

i first added it as a public var. but then noticed that it maybe should come from the platform secrets. maybe we can add this as a two stage value that can be overridden by the user?
but what should be the priority?
if nothing is configured in the descriptor it comes from the platform? and if nothing is in the platform what then? 8.8.8.8 or should it be the bastion internal ip ?

[alvarocurt]

Hi @bjoern-r
You are right. The variable priority is set on file load_variables.yaml.
Basically:
Site variables < Private variables < Public variables

If a variable is not defined in any of those places, it is up to the component's developer to decide if the pipeline should fail or have a different default value.

I agree on having the 5G-Core DNS as both a site and a public variable. Also I doesn't mind if you set the bastion or an external resolver as the default DNS. Bastion should have less latency but we may put too much pressure on it.

Your code is perfectly valid then! Thank you so much!

Solved