Distributed vPoller

[marguz]

Please move this if Im asking in the wrong forum.

I'm looking into leveraging Zabbix, Zabbix Proxies, vPoller(s) for an environment that is currently 435 ESXi hosts and 6000+ VMs over two datacenters.

My problem is understanding the deployment of the vPoller. I know this is a Zabbix forum, but I cannot find a vPoller forum to post this question...

vPoller is distributed and so I'm assuming I can "just" install the vPolle Worker(s) on seperate VMs (seperate from the vPoller Proxy and Zabbix Proxies). How do these vPoller Workers know where to pump the collected data to and how does the vPoller Proxy know what vPoller Workers are out there?

I compiled the Native vPoller module, so does that mean I need to have the vPoller Workers on VMs that have the Zabbix (Server, Proxy, Agent) process running?

I was thinking I could just have as an example this, each one being a VM:
1 - Zabbix frontend
1 - Zabbix DB
1 - Zabbix Server
2 - Zabbix Proxies
1 - vPoller Proxy
2 - vPoller Worker

I just dont get how does the vPoller proxy understand there are many workers and does the Zabbix Server only care about the vPoller Proxy?

Also the data flow:
Does the vPoller Worker collected data get passed to the vPoller Proxy then to the Zabbix server or do they send directly to the Zabbix Server?