Has this bug (http://openvswitch.org/pipermail/discuss/2015-July/018242.html) ever been found & fixed?
If not, using OVS in this scenario will still break your cluster horribly.
It causes no problem whatsoever with a single non-clustered host.
Also, AFAIK, OVS *still* doesn't implement an IGMP querier, so if you were relying on the Linux Bridge IGMP querier (like I am), you're screwed there, too. (My switches support IGMP snooping, but my router doesn't have an IGMP querier.)

I guess I'm not upgrading to 4.x anytime soon :-(.

-Adam


On 16-05-06 04:21 AM, lyt_yudi wrote:
hi

recommend you use this mode:

http://pve.proxmox.com/wiki/Open_vSwitch 
<http://pve.proxmox.com/wiki/Open_vSwitch>

Example 2: Bond + Bridge + Internal Ports


在 2016年5月6日,下午4:59,Laurent CARON <[email protected]> 写道:

Hi,

I'm running a 3.4 cluster and am planning the upgrade.

It seem the vlan package will be removed upon upgrade (vlan is in conflict with 
pve-manager).

I've got the following /etc/network/interfaces config (on proxmox hosts):

auto bond1
iface bond1 inet manual
        slaves eth8 eth9
        bond-miimon 100
        bond-downdelay 200
        bond-updelay 200
        bond-mode 802.3ad

auto vlan34
iface vlan34 inet manual
        vlan-raw-device bond1

auto vmbr34
iface vmbr34 inet manual
        bridge_ports vlan34
        bridge_stp off
        bridge_fd 0

The ports on the switch are trunk ports carrying all VLANs used by my VMs.

I'm then using vmbr34 interface as a network interface on my hosts.

Will it be supported with proxmox 4 (since the vlan package will be removed) ?

Thanks
_______________________________________________
pve-user mailing list
[email protected]
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
_______________________________________________
pve-user mailing list
[email protected]
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

_______________________________________________
pve-user mailing list
[email protected]
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Reply via email to