Re: oVirt 4.3.5.1 failed to configure management network on the host

LACP supports works with 2 switches, but if you wish to aggregate all links - you need switch support (high-end hardware). Best Regards, Strahil NikolovOn Aug 6, 2019 18:08, Vincent Royer <vincent@epicenergy.ca> wrote:
I also am spanned over two switches. You can use bonding, you just can't use 802.3 mode.
I have MGMT bonded to two gig switches and storage bonded to two 10g switches for Gluster. Each switch has its own fw/router in HA. So we can lose either switch, either router, or any single interface or cable without interruption.
On Tue, Aug 6, 2019, 12:33 AM Mitja Pirih <mitja@pirih.si> wrote:
On 05. 08. 2019 21:20, Vincent Royer wrote:
I tried deployment of 4.3.5.1 using teams and it didn't work. I did get into the engine using the temp url on the host, but the teams showed up as individual nics. Any changes made, like assigning a new logical network to the nic, failed and I lost connectivity.
Setup as a bond instead of team before deployment worked as expected, and the bonds showed up properly in the engine.
ymmv
I can't use bonding, spanned over two switches. Maybe there is another way to do it, but I am burned out, anybody with an idea?
The server has 4x 10Gbps nics. I need up to 20Gbps throughput in HA mode.
Thanks.
Br, Mitja

On 08. 08. 2019 11:42, Strahil wrote:
LACP supports works with 2 switches, but if you wish to aggregate all links - you need switch support (high-end hardware).
Best Regards, Strahil Nikolov
I am aware of that. That's why my idea was to use bond1 (LACP) on eth1+2 on switch1 and bond2 (LACP) on eth3+4 on switch2 and then team together bond1 + bond2. With this config theoretically I should get bonding spanned over two switches. Technically it worked, redundancy and aggregation. The problem was deploying self-hosted engine, because the script was unable to configure management network. If I use bonding spanned over two switches as you suggest, based on documentation https://access.redhat.com/documentation/en-us/red_hat_virtualization/4.2/htm... my options are: - Mode 2 (XOR policy) - Mode 5 (adaptive transmit load-balancing policy): no use of bridges - Mode 6 (adaptive load-balancing policy): same limitation of mode 5 Basically only Mode 2 looks usable for us. Regards, Mitja
participants (2)
-
Mitja Pirih
-
Strahil