Re: Illumos and bhyve
- In reply to: Chris Moerz : "Re: Illumos and bhyve"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sat, 28 Dec 2024 12:58:15 UTC
Thanks Chris. Yea I noticed the rust and wondered if that would be an issue. Seems as though there may be more straightforward ways to achieve some of the manageability goals. But if others think the oxide pathe makes sense, please chime in On Sat, Dec 28, 2024, 5:16 AM Chris Moerz <freebsd@ny-central.org> wrote: > Hi Greg, > > On Fri, 27 Dec 2024, Greg Wallace wrote: > > > Hi everyone, > > > > In reviewing the issue tracker ( > https://docs.google.com/spreadsheets/d/1pM3boLIcKj26lq7XbwGMBXW0-F84We3r_Xc3zovHjaQ/edit?usp=sharing > ) > > > > I spotted the update in cell M18. (One all sidenote about spreadsheet > hygiene, when adding an update, please include your name.) > > > > Is anyone interested in doing a review of the linked repo to see what it > would entail to bring Illumos bhyve management code into bhyve? I know > they've put > > significant effort into live migration for instance. > > I did check the user space daemon by 0xide a while back: > https://github.com/oxidecomputer/propolis > > That's the user space component that runs on top of the Illumos based > bhyve they're running. As far as I could tell back then: it's based on > Rust, which is great in terms of marshalling and unmarshalling data when > sending machine state across a network. Porting that to FreeBSD (i.e. C) > is going to be a major undertaking. > > A few months back there was a very lively conversation about bringing Rust > into base. It was even discussed during the developer summit, if I > remember correctly. I do remember that distinctly because I wasn't there > in person but watched the live stream and would have loved to raise bhyve > as a side benefit, when it was brought up. I do understand how much effort > rust in base would create though. > > We might be able to patch up bhyve's kernel module to allow 0xide's rust > daemon from ports. I'd be happy to be proven wrong (in the positive > sense), I should add. > > Thia is still going to be considerable work, I imagine. Maybe it would be > worth a GSoC project? > > I could run through a code review on 0xide and FreeBSD's kernel code to > get a better understanding on the differences, though I imagine there's > people better suited to do this (i.e. with more kernel experience), and > who would be able to pull this off more quickly than me. > > chris > > > > > If there are any takers, please reply on here or to me so we can map out > a plan. > > > > Thanks! > > > > ---Greg Wallace > > Director of Partnerships > > [images?q=tbn:ANd9GcTcCtGVcNyNgZPByRISaVlEZNSxcr9UK16TFA&s] > > Schedule a meeting > > netactuate.com | anycast.com > > M +1 919.247.3165 > > linkedin.com/in/gtewallace/ > > > >