URL:
Reporter RHNID:
Section: -
Language: en-US (English)
Workaround:

Description: This exercise intends to demonstrate that with the 'activebackup' runner, the loss of the active port will cause the secondary interface to take over instantly without any disruption to ongoing network traffic. 

In step 10 on p.59 we  issue 'nmcli con down team0-port1' at which point eth2 becomes the active port. teamd does the right thing, but unfortunately traffic is still disrupted and a continuous ping in a 2nd terminal will only see packet loss after the failover.

Step 11 says "Notice that the backup port is now active and receives all the traffic." Unfortunately, all I see is echo requests going out but no echo reply being received.

Here is the setup:

servera VM:

eth1  - connected to privbr1 on the kvm host

eth2 - connected to privbr2 on the kvm host

team0 using the activebackup runner (192.168.0.100/24)

serverd VM: (configured by the lab script)

eth1 - connected to privbr1 on the kvm host (192.168.0.254/24 no teaming)

As long as the active port on servera is eth1, I can see traffic flowing correctly . The moment I disconnect eth1, and eth2 becomes the active port, I get 100% packet loss.

My understanding is that privbr1 and privbr2 behave like two isolated VLANs. We are not dealing with a single broadcast domain here. So I think our classroom design is broken. The interfaces used for teaming (eth1 and eth2) of all the VMs should be plugged into the same private bridge. Then the activebackup failover would actually have a chance to work correctly. We need a single broadcast domain.