Illumos/OmniOS support

[BobTB]

How much work would it be to support his for OmniOS which is running bhyve? Would be a great way to get more users.

[yaroslav-gwit]

It could be an interesting and challenging task to implement hoster under Illumos/OmniOS, but I am afraid I know very little about the system itself.

So I'll have to look into "the Illumos' way of doing things" first. Also, it would be nice to have someone from the community guide me through a few things if I get stuck.

Other than that, I'll be happy to collaborate on this further, as long as there is at least some interest from other people.

[yaroslav-gwit]

I tried playing with OmniOS a week back and it looked really cool, but if I was to port the code the underlying HosterCore implementation will have to be much different:

• we'll have to ditch the vm_supervisor service, because OmniOS has it's own implementation of it
• ZFS features are not 100% on par with FreeBSD 13.2, which means you can still replicate the VMs across the systems, but if you wanted to re-install the system (from FreeBSD to OmniOS) and re-use the pools - you might not be able to import them
• PF is not part of the system, so our future integration (HosterCore+PF) would have to be thrown away in this new code base
• I am developing and testing HosterCore on ZimaBoard computers, and it looks like the OmniOS has issues running (installing?) on top of that platform, so unless there is some spare hardware I can find that is compatible with OmniOS - I wouldn't be able to start the porting process
• If there is something similar to HAST, please point me to it, because I am planning of using it for the hyper-converged setups with synchronous data-links between the hosts
• VNC connections, and the serial console connections are managed differently from FreeBSD's implementation, so I'll have to account for that as well

I want to keep the CLI and REST API calls the same (or at least as close as possible) across both systems, so even if the underlying implementation is a bit different - there is still a common ground for the future WebUI. The idea is to make the REST API implementation so similar, that the end user would not notice any difference in the WebUI/CLI.

Another cool thing is the availability of the KVM Zones, which could help us achieve the Live VM migrations badge, and an easier migration path for the VMWare users (bhyve isn't really the best fit there, because it doesn't support some older BIOS-based systems).