ML
    • Recent
    • Categories
    • Tags
    • Popular
    • Users
    • Groups
    • Register
    • Login

    Network setup for production KVM

    IT Discussion
    kvm networing setup
    8
    18
    2.4k
    Loading More Posts
    • Oldest to Newest
    • Newest to Oldest
    • Most Votes
    Reply
    • Reply as topic
    Log in to reply
    This topic has been deleted. Only users with topic management privileges can see it.
    • FATeknollogeeF
      FATeknollogee @stacksofplates
      last edited by

      @stacksofplates No loss of access.
      But that is me creating a bridge from my Fedora desktop, the bridge is on a remote server

      1 Reply Last reply Reply Quote 0
      • stacksofplatesS
        stacksofplates @JaredBusch
        last edited by stacksofplates

        @jaredbusch said in Network setup for production KVM:

        So for those who have KVM in production, how do you setup the network?

        In Hyper-V i always team the NICs in switch independent mode and then make the vSwitch on the team. The host will have access to the guest VM networks.

        For my home lab (Fedora 26) and on my desktop (F25) and laptop (F26) I just use the macvtap in bridged mode. But I have no host to guest communication. This is not an issue for my lab or desktop. But I do not want this in production.

        0_1510330763991_92fdf487-e62e-412d-967c-c8583e5fe2f4-image.png

        So if I have 2-4 NICs in a server, assuming Fedora 26 or RHEL 7:

        • How should I team them?
        • Should I create a bridge?
        • What source mode should I use?

        I have stuff in prod running with macvtap and then there is a separate network for access to the host if the VM needs it. But I also have stuff in prod with a bridge and full access. Just depends on what you want.

        IIRC NetworkManager doesn't play nicely sometimes with libvirt and other bits (or it may have just been with bridging in general). You can build a bridge or bond directly from Virt-Manager as well, but I believe it uses network and not NetworkManager.

        0_1510337500562_bond.png

        1 Reply Last reply Reply Quote 1
        • ObsolesceO
          Obsolesce
          last edited by

          Doesn't using a bridge have a negative impact network performance of the VM?

          JaredBuschJ 1 Reply Last reply Reply Quote 0
          • JaredBuschJ
            JaredBusch @Obsolesce
            last edited by

            @tim_g said in Network setup for production KVM:

            Doesn't using a bridge have a negative impact network performance of the VM?

            Not usually.

            1 Reply Last reply Reply Quote 0
            • M
              marcinozga @JaredBusch
              last edited by

              @jaredbusch why don't you add host interface to macvtap bridge and route all traffic through it? I'm doing that with my LXD containers and host.
              Here's how to do it:
              http://noyaudolive.net/2012/05/09/lxc-and-macvlan-host-to-guest-connection/

              JaredBuschJ 1 Reply Last reply Reply Quote 0
              • JaredBuschJ
                JaredBusch @marcinozga
                last edited by

                @marcinozga said in Network setup for production KVM:

                @jaredbusch why don't you add host interface to macvtap bridge and route all traffic through it? I'm doing that with my LXD containers and host.
                Here's how to do it:
                http://noyaudolive.net/2012/05/09/lxc-and-macvlan-host-to-guest-connection/

                All you are doing there is making a bridge on the host.

                Also LXC is containerization not virtualization..

                scottalanmillerS M 2 Replies Last reply Reply Quote 0
                • scottalanmillerS
                  scottalanmiller @JaredBusch
                  last edited by

                  @jaredbusch said in Network setup for production KVM:

                  @marcinozga said in Network setup for production KVM:

                  @jaredbusch why don't you add host interface to macvtap bridge and route all traffic through it? I'm doing that with my LXD containers and host.
                  Here's how to do it:
                  http://noyaudolive.net/2012/05/09/lxc-and-macvlan-host-to-guest-connection/

                  All you are doing there is making a bridge on the host.

                  Also LXC is containerization not virtualization..

                  It's both. Containerization is Type-C Virtualization. It's always been considered a form of virtualization, even though it is a totally different technological approach.

                  1 Reply Last reply Reply Quote 0
                  • scottalanmillerS
                    scottalanmiller
                    last edited by

                    Containers are the trendy new term for OS Level Virtualization. https://en.wikipedia.org/wiki/Operating-system-level_virtualization

                    1 Reply Last reply Reply Quote 0
                    • M
                      marcinozga @JaredBusch
                      last edited by marcinozga

                      @jaredbusch no, you're creating macvlan interface on physical host adapter. And by routing traffic through it, you allow host to communicate with guests.
                      Containers or VM guests makes no difference here.

                      Reid CooperR 1 Reply Last reply Reply Quote 0
                      • M
                        marcinozga
                        last edited by

                        Perhaps this explains it better: https://superuser.com/a/368023

                        1 Reply Last reply Reply Quote 0
                        • Reid CooperR
                          Reid Cooper @marcinozga
                          last edited by

                          @marcinozga said in Network setup for production KVM:

                          Containers or VM guests makes no difference here.

                          Rarely does.

                          1 Reply Last reply Reply Quote 0
                          • 1 / 1
                          • First post
                            Last post