Special thanks to Carlo Cianfarani for providing this post.

In this post we will explore how to troubleshoot and fix an issue on EVE-NG involving Linux VMs connected through a virtual router instance (Cisco XRd) implemented as docker container. Some time ago, indeed, I managed to extend EVE-NG to support docker containers.

Lab scenario

In the scenario depicted, Linux VM ethernet interfaces (e.g. plain e1000 or paravirtualized virtio-net) are attached to two different IP subnets, each implemented by a Linux bridge. Linux VMs run on QEMU hypervisor which, at userspace level, starts a multithreaded process for each VM. Essentially, there is a main thread plus additional threads, one for each VM’s assigned vCPU.

We can check how EVE-NG networking is implemented under the hood by using:

In the above vnet0_1 and vnet0_2 are both Linux bridge implementing ethernet links connecting the VMs to the virtual router’s interfaces. vunl0_1_0 and vunl0_2_0 are Linux TAP network devices implementing the VM virtual ethernet interfaces’ backend.

vnet0_3_1 and vnet0_3_2 on the other hand are Linux veth network devices where the paired veth lives inside the docker container’s network namespace.

The following command enters the XRd container’s namespaces (PID, network and User) and confirms the associations between veth pairs.

Continue reading the post on Patreon .