Hi
After running netdata on aks and configuring type-LoadBalancer, I had a problem with the available metrics. If I open the netdata dashboard, I have access to the basic metrics. If I choose one of the netdata-child dashboards, I have access to full metrics for all pods, applications.
However, if I provide metrics on port 19999/api/v1/allmetrics for prometheus, I only have access to basic metrics (those from the basic netdata-parent dashboard). Statds metrics are visible. Is it possible to share also netdata-child metrics on port 19999 for prometheus? (I’m talking about cpu resource usage statistics, memory by pods) Is additional configuration required on the netdata or prometheus side?
Hi @Kamil
You are correct, /api/v1/allmetrics returns only metrics collected by the instance, if you want to gather children metrics from the parent instance:
/api/v1/allmetrics?format=prometheus_all_hosts
This same metrics was available.
How did you check they are same?
Parent-2children setup
0 ~ $ curl -s http://X.X.X.X:19999/api/v1/allmetrics?format=prometheus | wc -l
56
0 ~ $ curl -s http://X.X.X.X:19999/api/v1/allmetrics?format=prometheus_all_hosts | wc -l
4188
Lets see what you get, can you change X.X.X.X to you netdata’s ip address or domain name and execute those requests?
Just to give some context here, wc is a command-line tool that outputs statistics about it’s input.
Here we are giving as input the JSON text that is returned by curl and the flag -l which means that we want to count the number of lines.
The increase in the number of lines indicates that the returned JSON is considerably larger, as it contains all the metrics from all the children.
Thanks @ilyam8 for providing such an easy proof.
1 Like