2 Leaf 2 Spine with Cumulus VX not working if tutorial is followed (VMware Workstation)

I followed the Instructions on creating a leaf-spine archticture with 2 leafs and 2 spine switches closely and created 4 cumulus VX vms (on VMware Workstation). I imported the OVF once and then did changes to the hostname and ip address, edited interfaces and quagga configuration and then cloned them until i had 4 machines. after the cloning i changed the interfaces and quagga config files for every vm according to the tutorial. one issue i see is that the tutorial states that leaf 1 should have a 24 bit subnet on swp3, while the leaf2 should have a 25 bit subnet on swp3. that does not make much sense (but even when i correct the leaf 2 config to have a 24 subnet it does not work properly).

after having all configurations according to the tutorial (https://community.cumulusnetworks.com/cumulus/topics/creating-a-two-leaf-two-spine-topology-using-cu...) and rebooting all of the vms to make sure all necessary services are restarted the ping from leaf1 will not reach leaf2.
the tutorial states to put the first port to dhcp, that means it will get not only an ip adress but also a default gateway.

the routes on leaf 1 look like this:

default eth0 * swp3 * eth0

In this example the network is managed by DHCP and the is the default gateway

how on earth can a ping to leaf 2 ( ever work if the routes would tell the system to send the packet to the default gateway? even if i delete the default route the tutorial setup will never ever send the packet towards swp3 (towards the spine) to forward the packet to leaf2.

i think there are some steps missing.

since it is possible to package several VMs as one OVA, you guys probably want to make it easier for everyone to test this by just packaging 4 VMs (2leaf & 2 spine vms) into one package/OVA and let the people test your working setup.

thanks in advance

4 replies

Userlevel 5
Jan, Can you share more of your configuration? What is your interfaces setup?
refer to this thread for the solution:

Userlevel 4
Jan I saw a ticket (@Scott the ticket is UD-563) float by on this yesterday.@Jan -- There is an action item to correct the issues mentioned in your initial e-mail. We are listening and we'll get this corrected ASAP. We have a ton of additional content to add around VX as well, just need to find the cycles to prepare it for consumption but to give you a sneak peak, we're talking about libvirt and KVM for large scale simulations (450+ nodes) and using vagrant in some more complex examples to build the interconnected VMs for you. We're doing all this stuff internally but are moving at hyper-speed and as is always true, documentation can be the hardest task. THANK YOU FOR FINDING THESE ISSUES!
thank you. it is always good to see a company so responsive.
keep up the good work.