Playing around with my lab at home I have 2 nested hosts running on the same server and then I have 1 physical host on a separate stand alone hardware/server. The physical host has LACP enabled on the switch and I configured a port group to use LACP. So the LACP enabled host seems to be working fine. The 2 nested hosts are running without LACP as workstation doesn't support LACP (the host these hosts are running under). They are in their own port group. So I have 2 different port groups, 1 PG for LACP and 1 for non-LACP hosts.. SO the setup seems fine and I can migrated from the LACP host but when I do the guests I migrate stop pinging.
Is there a way to make vmotion of guests to work from a LACP enable PortGroup to a non-LACP port group WITHOUT having to go into the settings and select the "correct" port group in the guests settings?? When I go in and manually select the correct port group the guest is pingable again, but if I migrate back it again stops pinging until I select the correct port group.
thanks,
Mike
IMO this is not an issue with LACP per se, but with the port group configuration. When you vMotion a VM (prior to vSphere 6.0, where the destination port group is selectable) it will be connected to a port group with the exact same name as the one on the source host. It's up to you to configure the network, so that the port group provides the same connectivity.
André
ok but in this case I need to have 2 port groups defined in the same dvswitch so unfortunately (as far as I know, correct me if I'm wrong) they can't share the same name...
So understanding the setup and limitations are you able to supply a work around or suggestion?
thanks,
Mike
You are right, a port group name is unique, so the only way to solve this is to ensure that the port group provides the same connectivity to the physical network in means of tagged/untagged port groups. With a mix of a physical host and virtual hosts that may be tricky.
André
Really I don't know if this is a nested/virtual and physical combo limitation.. Ultimately this could be any type of host but the network config is the limiting factor.. so I'm guessing if I had any number of physical hosts only but only 1 of them supported LACP in the cluster that I would still run into this limitation.
I guess an enhancement would allow you to specify a destination network per host so that if any guest was migrated to that host it would default to the port group of that host selected by default.. Doesn't seem like too crazy of an idea if you ask me I think this differs from the ESXi implementation that you described above in that the guest is allowed to pick the destination (in my fix the host would have the default group, not the guest).
That said, until that is implemented I guess I basically have 1 option.. to disable LACP on my physical host and share 1 port group... darn shame!
thanks,
Mike
Hi techno10,
Another option is to script the virtual machine port group change and vMotion migration (i.e. via PowerCLI). Alternatively, you may consider adding additional physical NICs to support this (even USB would be fine, if performance is not critical).