I have 32gb of ram so not concerned about resource usage. I miss how easy it was to manage and set up dockers in Coral. Someone who's never used docker before could do it without documentation.
I wonder when docker containers themselves can be set up in the FreeNAS UI without the useless rancher interface?
So far FreeNAS 11.1. has only provided a "user friendly way" to download and start a Docker VM based on rancherOS to acts as a docker host. FreeNAS docs do show you how to get the rancher-server going, but I agree these docs are minimal and there are things you'd probably want to configure on the docker host before pulling and running the rancher-server and these are not covered. The FreeNAS docs may even give the mistaken impression that the use of the rancher-server is obligatory, it is not. You could use portainer for example as a way to manage your docker containers. But you'd still need to do some work at the CLI to set this up.
I think most people would argue that basing a "Docker VM" on rancheros rather than boot2docker, which was used in Corral, is technically superior and is far from useless. But rancheros is kind of "Linux, but not as we know it Jim" , so having to configure the rancheros based Docker host might be a challenge and not everyone will have the inclination to get to grips with the documentation at
http://rancher.com/docs/os/v1.1/en/configuration/ etc.
One major piece of the puzzle that the FreeNAS docs gives no hint about is how docker containers running in a Virtual machine may access data in your zpool. If the FreeNAS devs decided to put all their docker eggs in the Rancher basket, you'd think they might have offered some guide/advice here. Obviously it must be done by network shares, but what is the best method? Did they assume you'd use the Rancher-NFS service available if you installed the rancher-server and configured via the RancherUI? There are alternatives.
As mentioned above, those familiar with Linux may have already taken the easy route of using docker-ce and docker-compose in their preferred distro by just using a traditional network mount which then maps to docker volumes as necessary. Their reaction to the new "Docker Vm" function in FreeNAS 11.1 maybe, so what?
For apps that need to access zpool data, jail performance will be better than dockerized apps but then not all apps can run in jails and some are just far easier to use as docker containers. Of course both jails and VMs consume resources, but I don't think it is correct say allocating CPUs to a VM results in "Now you just locked FreeNAS out of most of your CPU". As far as bhyve is concerned my basic understanding is that a virtual "CPU Core" roughly equates to a kernel-thread which then gets schedule across as many real cores as you have on your machine.
