Provide instructions on how to run with a parent outside the k8s cluster

[cakrit]

We have seen the need to have the parent outside the cluster and are doing it ourselves. We need to be explicit about this being a very valid option and explaining which settings will allow this to happen.

[ilyam8]

Our setup is children->in-cluster parent-> outside-cluster parent.

Another option is children -> outside-cluster parent (no in-cluster).

Do we need both? In any case, it is not just an option (instruction) but should be implemented.

[cakrit]

children->in-cluster parent-> outside-cluster parent can work very well, but we'll then probably need to say how to configure the in-cluster parent to ensure it uses very limited resources. Does the in-cluster parent still need persistent storage in our setup, in order to prevent data loss?

[ilyam8]

@cakrit

• We use the outside-cluster parent because it is easier to debug/catch problems (crashdump), and easier to run a custom branch (better for developing).
• We use the in-cluster parent because (from what I heard) it is the default we suggest to use so we need to use it ourselves.

---

persistent storage in our setup, in order to prevent data loss

Yes.

very limited resources

Our in-cluster parent is opposite to it.

[Rezrazi]

I'm not sure that this issue describes what I'm trying to achieve:

• I have a cluster on which I've installed the netdata chart and everything is running smoothly
• I have another cluster, from which I want to collect Kubernetes metrics and report them to my initial setup

Any indication on how a setup like this can be achieved?

@ilyam8