VMware Cloud Community
vSohill
Expert
Expert
Jump to solution

VMware Load-Based Teaming (LBT) and Route Based on Physical NIC Load

Hi,

I came across a question regarding LBT, IF I am using Route Based on Physical NIC   Route Based on Physical NIC Load for a portgroup with 2 Uplinks, Each vmnic connected to different physical Switch for redundancy. If  VM01 sends a traffice via vmnic0 to external switch S01, then vmnic0 went down. VM01 will send the traffic via vmnic1 that is connected to physical switch  S02. The Notify Switches  opetion will tell the Switch S02  about  the new MAC  what about Switch S01 that lost the contact with vmnic0. How this switch will be notfied to remove the MAC and the MAC will not be presented on the two Switches

Thank you

1 Solution

Accepted Solutions
daphnissov
Immortal
Immortal
Jump to solution

The behavior is essentially the same. If the link is down, the VPC peer shows the port down as well. When gratuitous ARPs are then received on the other VPC peer, the table has a more recent time than the other member and takes precedence. Packets are then switched out of the corresponding port on the other peer. Also keep in mind that with LBT, the process of reassigning VMs to other uplinks occurs naturally as part of the algorithm. Every 30 seconds a sampling is taken of both uplinks and their utilization. If any one is greater than 75% utilized, a VM may be reassigned to another uplink. In that case, RARPs are still sent to the upstream switch signaling that MAC is now accessible in another location rather than waiting for the VM to initiate traffic thereby reducing any switching convergence time.

View solution in original post

5 Replies
daphnissov
Immortal
Immortal
Jump to solution

If S01 is still functional and available from S02, when ESXi notifies S02 it will brodcast RARP packets on behalf of VMs corresponding to each MAC. Because these packets have a broadcast address of FF:FF:FF:FF:FF:FF, any other switch in the same broadcast domain will get this information and will update their MAC-to-port tables accordingly.

hussainbte
Expert
Expert
Jump to solution

How does the situation differ if you have a VPC across 2 Nexus 5K switches with same load balancing policy..?

If you found my answers useful please consider marking them as Correct OR Helpful Regards, Hussain https://virtualcubes.wordpress.com/
vSohill
Expert
Expert
Jump to solution

Yes I do have the same question .

0 Kudos
daphnissov
Immortal
Immortal
Jump to solution

The behavior is essentially the same. If the link is down, the VPC peer shows the port down as well. When gratuitous ARPs are then received on the other VPC peer, the table has a more recent time than the other member and takes precedence. Packets are then switched out of the corresponding port on the other peer. Also keep in mind that with LBT, the process of reassigning VMs to other uplinks occurs naturally as part of the algorithm. Every 30 seconds a sampling is taken of both uplinks and their utilization. If any one is greater than 75% utilized, a VM may be reassigned to another uplink. In that case, RARPs are still sent to the upstream switch signaling that MAC is now accessible in another location rather than waiting for the VM to initiate traffic thereby reducing any switching convergence time.

vSohill
Expert
Expert
Jump to solution

Perfect, Thank you

0 Kudos