Scrap all /metrics endpoints pods from DaemonSet, like Prometheus (Pod&ServiceMonitor)

[Momotoculteur]

Hello Guys
I have this configuration defined in my image :

Multiple nodes which run in each :

• One vector agent
• One instance of cAdvisor

cAdvisor is like kube-state-metrics, expose a lot of metrics on 8080/metrics. I can scrap directly KSM via prometheus_scrape Vector Datasource which work nicely, but this is a replicatSet.
But cAdvisor run in daemonset to have node base metrics, not like KSM. I thought that i could make the same with cAdvisor, to call directly the pod and do not pass to the service :

     sources:
      cadvisor:
        endpoints:
        - http://cadvisor:8080/metrics
        scrape_interval_secs: 60
        type: prometheus_scrape

But this seems to pass thought the service, because i have a lot of metrics, but not 100 % from all my nodes :wat:

I'm a bit lost because i'm not expert in Kube & Vector, my apologies for this idiot question.

Do i have to use a Prometheus agent, and scrap my pods with Pod&Service monitor, and then be scrapped from Vector with prometheus exporter source ? Or is this possible from Vector to scrap this kind of stuff directly without to use Prometheus ?

Thanks a lots guys

Ps: This is what i try to achieve