search

Gradiant / 5g-charts

Helm charts for 5G Technologies
Apache License 2.0
110 stars 43 forks source link

open5gs pods multiple replicas ( #90

Closed jmarkotic closed 1 year ago

jmarkotic commented 1 year ago

hello, question, not an issue really. is it safe to scale pods to multiple replicas ? (default is 1 for all components) ? I can see comment in values.yaml that RWM mode is required if scale is required. Does this apply really to all components pods ? thanks, j

cgiraldo commented 1 year ago

Hi,

You will not be able to scale de deployments because we defined a HorizontalPodAutoscaler with maxReplicas to 1 in order to avoid people scaling the 5G functions.

If you need for example multiple amfs in your core, you have to deploy multiple open5gs-amfs, each one with a different name (for example playing with the helm release name), in order to register with a unique name in the open5gs-nrf.

If you are thinking in having multiple pods to have some sort of high availability for your 5G function, you can give it a try, but I really don't know the implications on the open5gs core. https://github.com/open5gs/open5gs/issues/1549 does not recommend it.

To scale a deployment you need to delete the corresponding hpa resource.

Please, let us know if you test it!

Kind Regards,

jmarkotic commented 1 year ago

Hi @cgiraldo, thanks for feedback. I was targeting for scaling and HA capabilities, hence having single service targeting 2 endpoints (like AMF). I tried scaling webui pod and got it to scale to 2 with LB service in front. Seemd to work but then again web must be easiest pod to scale and still work. I willl modify hpa to make it work with core functions. I will check that issue you refrenced and get back if i do additional testing.

DiogoCruz40 commented 1 year ago

@jmarkotic did you find any alternative?