Inband Management Examples Needed

[dankuednet]

An AVD example of how to configure inband management with and without out of band would really be helpful.

Problem1: When testing using ContainerLab and cEOS-lab, the Ma0 interface becomes unusable after an AVD configuration is downloaded.
a. It will not route out through the management interface.
b. Ma0 cannot be pinged even though it appears in "sh int".
c. This breaks the nameserver and radius configs because MGMT is unusable.

Problem2: I can't seem to be able to properly define in-band management.
a. Deploy will not let me use the same subnet as Ma0 although build will.
b. Trying to define vrf MGMT2 and a different subnet on a leaf switch port doesn't send a "no switchport" command in cfg to enable the port.
c. Defining MGMT2 vrf for in-band doesn't include routing.

[ClausHolbechArista]

Please provide some minimal example of the inputs you are giving and the expected config.
For inband management you need to use the inband_mgmt_* variables. Not the mgmt_ip.

[dankuednet]

Problem1: Attached is an excerpt of our ContainerLab config, the default/pre-deploy tests/results for spine1 and leaf1, the post-deploy results for leaf1. This test was designed to only use OOB Management and no Inband yet.

I believe that solving the OOB routing issue may be the first step in resolving the Inband issue.
Looking at the results after deployment, you can see that the Docker Leaf1 route has changed and so has the cEOS Leaf1 route. The docker lab has lost the gateway for Leaf1. The cEOS has lost both the static route and the direct connected route to the docker lab network.

Leaf1_default_config_and tests.txt
Leaf1_deployed_config_and_tests.txt
Spine1_default_config_and_tests.txt
ContainerLab_config_excerpt.yml.txt

[ClausHolbechArista]

You have not included the AVD inputs here, nor the configuration for leaf1, so it is difficult for me to evaluate the outputs. The default AVD OOB config would place the interface in the VRF MGMT, so you need to remember to give the vrf context on every show-command and ping. To me it looks like you are testing in VRF "default".

[dankuednet]

I believe the attached files are the ones I was using.
UED_avd_2025-06-02.zip

Sorry for the delay. Git, VSCode, and Docker are not playing nicely. Git was not syncing properly, after an update VSCode isn't communicate with container lab (docker network issue), and VSCode won't open the docker development container.
bang head on desk

[ClausHolbechArista]

Looking at your inputs here, indeed the management interface is in the MGMT VRF, so you have to do your testing with that context, like show ip route vrf MGMT. A nice EOS feature is cli vrf MGMT where it changes the context of all show commands to that VRF. Remember to switch back when you want to look at VRF "default". Ref. https://arista.my.site.com/AristaCommunity/s/article/vrf-interaction-using-eos-and-bash

[dankuednet]

Thanks for enlightening me on the context change. I believe that solves Problem 1.