so the container is to provide access to nfs-client when not locally available on the host?
RancherOS does not have a nfs-client available when you first set it up. Since RancherOS has more or less everything containerized, the nfs client need to be loaded as a container so the OS can make use of it, i do this in the yml file (its like installing a package for the os)
PS: i am not an expert on RancherOS, this is just a first approximation for me after quickly skimming the extensive wiki
To my setup:
Terminology:
rancher -> GUI for setting up containers
RancherOS -> OS
rancher:agent -> container to give rancher administration rights on docker-machine.
RancherOS+rancher+rancher:agent -> bhyuveVM -> 192.168.0.16
UbuntuVM+rancher:agent ->192.168.0.17
The installed rancher:agent allows rancher to administer the given host. So i can administer as many docker machines as i want. So if you want to have certain machines that run at certain times you can do this from the rancher GUI, no problem.
What i am hoping for is that at somepoint (freenas 11.1 or 11.2) will allow setting up RancherOS through the freenas GUI so i can setup a couple of rancherOS-VMs via GUI and have one rancherOS just to administer the other VMs, basically a HA setup.
hoping to move everything but Plex into containers
This is exactly what i have done (except replace plex with emby)
These are just the containers for my RancherOS system, i have them scheduled to restart once a week, or depending on the container once a day. Some of them i start by hand some of them i have set to inactive because i wanted to try out alternatives, simple export of docker-compose and rancher-compose files makes it super easy to move from one host to another.
unrelated:
I recently moved away from HA-proxy on my freenas to use HA-proxy on my pfsense firewall the 200$ extra for the pfsense which has a 100% update time was worth every penny :D and seperates a littlebit the load.