Re: [lxc-users] Bonding inside container? Or any other ideas?
Hi Andrey, Are you trying to use bond0 as the container's interface? If so, I think that's going to cause issues. You need an interface behind (in front?) bond0. Here is my interfaces - you're going to need some device between bond0 and LXC, though. lai@R610-LXD1-Dev-DMZ:~$ cat /etc/network/interfaces # This file describes the network interfaces available on your system # and how to activate them. For more information, see interfaces(5). source /etc/network/interfaces.d/* # The loopback network interface auto lo iface lo inet loopback auto eno1 iface eno1 inet manual bond-master bond0 auto eno2 iface eno2 inet manual bond-master bond0 auto eno3 iface eno3 inet manual bond-master bond0 auto eno4 iface eno4 inet manual bond-master bond0 auto bond0 iface bond0 inet manual bond-mode 4 bond-slaves none bond-miimon 100 bond-lacp-rate 1 bond-downdelay 200 bond-updelay 200 bond-xmit-hash-policy layer2+3 auto br0 iface br0 inet static bridge_ports bond0 bridge_maxwait 10 address 10.1.1.139 netmask 255.255.0.0 broadcast 10.1.255.255 network 10.1.0.0 gateway 10.1.1.5 dns-nameservers 10.1.1.84 8.8.8.8 From: "Andrey Repin" <anrdae...@yandex.ru> To: "lxc-users" <lxc-users@lists.linuxcontainers.org>, "lxc-users" <lxc-users@lists.linuxcontainers.org> Sent: Wednesday, November 22, 2017 5:25:11 PM Subject: Re: [lxc-users] Bonding inside container? Or any other ideas? Greetings, Lai Wei-Hwa! I'm not sure I follow. I have multiple servers running Bond Mode 4 (for LACP/802.3ad). 802.3ad (mode 4) requires switch support. Unfortunately, my switch is "managed", but does not offer this essential specification. BQ_BEGIN I then created a bridge, br0 which becomes the main (only) interface. BQ_END After having a hard time with some of the configurations, I avoid brctl like plague. It may be a tool to bridge physical interfaces, but for single host it is an extreme overhead. -- With best regards, Andrey Repin Thursday, November 23, 2017 01:20:52 Sorry for my terrible english... ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users
Re: [lxc-users] Bonding inside container? Or any other ideas?
Greetings, Lai Wei-Hwa! > I'm not sure I follow. I have multiple servers running Bond Mode 4 (for > LACP/802.3ad). 802.3ad (mode 4) requires switch support. Unfortunately, my switch is "managed", but does not offer this essential specification. > I then created a bridge, br0 which becomes the main (only) interface. After having a hard time with some of the configurations, I avoid brctl like plague. It may be a tool to bridge physical interfaces, but for single host it is an extreme overhead. -- With best regards, Andrey Repin Thursday, November 23, 2017 01:20:52 Sorry for my terrible english... ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users
Re: [lxc-users] Bonding inside container? Or any other ideas?
I'm not sure I follow. I have multiple servers running Bond Mode 4 (for LACP/802.3ad). I then created a bridge, br0 which becomes the main (only) interface. I'm using flat networking with no NATS between containers and edited the profiles to use br0. Everything works for me. I can't speak to the other bond modes, though. Thanks! Lai - Original Message - From: "Andrey Repin" <anrdae...@yandex.ru> To: "lxc-users" <lxc-users@lists.linuxcontainers.org> Sent: Tuesday, November 21, 2017 6:38:55 PM Subject: [lxc-users] Bonding inside container? Or any other ideas? Greetings, All! Some time ago I've managed to install a second network card into one of my servers, and have been experimenting with bonding on host. The field is: a host with two cards in one bond0 interface. A number of containers sitting as macvlans on top of bond0. Some success was achieved with bond mode 5 (balance-tlb) - approx 2:1 TX counts with five clients, but all upload is weighted on one network card. Attempt to change the mode to balance-alb(mode 6) immediately broke the loading of roaming Windows profiles, the issue immediately disappear once I switch back to mode 5. I suppose this happens because bonding balancer creates havoc with macvlan and own bonding MAC addresses, which the network can't easily solve, or Windows clients got picky and refuse to load stuff from randomly changed source. While I could turn back to internal LXC bridge and route requests between it and bond0 on host to dissolve the MAC issue, I'd like to see if there's a more direct solution could be found, such as creating a bonding inside container? Or if not, is there any other way to use bonding and maintain broadcast visibility range between containers and the rest of the network? -- With best regards, Andrey Repin Wednesday, November 22, 2017 02:23:22 Sorry for my terrible english... ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users
[lxc-users] Bonding inside container? Or any other ideas?
Greetings, All! Some time ago I've managed to install a second network card into one of my servers, and have been experimenting with bonding on host. The field is: a host with two cards in one bond0 interface. A number of containers sitting as macvlans on top of bond0. Some success was achieved with bond mode 5 (balance-tlb) - approx 2:1 TX counts with five clients, but all upload is weighted on one network card. Attempt to change the mode to balance-alb(mode 6) immediately broke the loading of roaming Windows profiles, the issue immediately disappear once I switch back to mode 5. I suppose this happens because bonding balancer creates havoc with macvlan and own bonding MAC addresses, which the network can't easily solve, or Windows clients got picky and refuse to load stuff from randomly changed source. While I could turn back to internal LXC bridge and route requests between it and bond0 on host to dissolve the MAC issue, I'd like to see if there's a more direct solution could be found, such as creating a bonding inside container? Or if not, is there any other way to use bonding and maintain broadcast visibility range between containers and the rest of the network? -- With best regards, Andrey Repin Wednesday, November 22, 2017 02:23:22 Sorry for my terrible english... ___ lxc-users mailing list lxc-users@lists.linuxcontainers.org http://lists.linuxcontainers.org/listinfo/lxc-users