Lab setup on macOS (OrbStack + containerlab + cEOS)

Everything you need to run the Course 2 cEOS lab on a Mac — no cloud, no nested virtualization. cEOS is a container, so a full 4-node VXLAN-EVPN fabric boots on your laptop in minutes.

✅ Validated on Apple Silicon (M-series) with OrbStack, cEOS 4.32.0F under Rosetta.

The stack: macOS → OrbStack (fast Docker + Linux VMs) → a Linux machine that runs Docker + containerlab → your cEOS nodes. Everything below runs inside the Linux machine, so your Mac stays clean.

graph LR
    MAC["macOS<br/>(your laptop)"] --> ORB["OrbStack<br/>Linux VM 'mylab'"]
    ORB --> DKR["Docker + containerlab"]
    DKR --> LAB["cEOS fabric<br/>clab-ceos-evpn-*"]
    classDef a fill:#1565c0,stroke:#90caf9,color:#ffffff,stroke-width:2px,font-weight:bold;
    classDef b fill:#2e7d32,stroke:#a5d6a7,color:#ffffff,stroke-width:2px,font-weight:bold;
    class MAC,ORB a; class DKR,LAB b;

---

Prerequisites

• macOS — Apple Silicon (M-series) or Intel. cEOS is amd64-only; on Apple Silicon it runs under OrbStack's built-in Rosetta emulation (automatic — no toggle to flip).
• A free Arista account (arista.com) to download the cEOS image (like Juniper's account for vJunos).
• ~8 GB free RAM for a 4-node fabric.

---

1. Install OrbStack

Download and install from orbstack.dev. OrbStack gives you a Docker engine and lightweight Linux VMs, far lighter than Docker Desktop or a full VM — and it does amd64 emulation automatically.

2. Create a Linux machine

containerlab manipulates Linux network namespaces, so it needs a real Linux host. Create one in OrbStack (this is your lab box):

orb create ubuntu mylab

(or use the OrbStack app → Machines → New). Then enter it:

ssh orb            # drops you into the machine → rwh@mylab:~$

Everything from here runs inside mylab.

Your Mac files are already mounted

OrbStack mounts your macOS home into the Linux machine, so a file you download to ~/Downloads on the Mac is visible inside mylab too — no scp needed.

3. Confirm Docker works

docker ps

You should get an empty table (not an error). OrbStack provides the Docker engine to the machine automatically.

4. Get and import the cEOS image

1. On arista.com → Support → Software Download → cEOS-lab, grab a recent stable build, e.g. cEOS64-lab-4.32.0F.tar.xz.
2. Import it (from wherever it landed — your mounted Mac ~/Downloads works):

docker import --platform linux/amd64 cEOS64-lab-4.32.0F.tar.xz ceos:4.32.0F
docker images | grep ceos

Apple Silicon: the --platform linux/amd64 flag is mandatory

cEOS binaries are x86_64. --platform linux/amd64 tags the image as amd64 so OrbStack runs it under Rosetta. Skip it and you'll get exec format error when the container starts. .tar.xz is fine — docker import decompresses it for you.

5. Install containerlab

bash -c "$(curl -sL https://get.containerlab.dev)"
containerlab version

6. Smoke test — one node (prove cEOS boots)

Before the full fabric, confirm a single cEOS node comes up on your Mac:

mkdir -p ~/ceos-lab && cd ~/ceos-lab
cat > smoke.clab.yml <<'EOF'
name: ceos-smoke
topology:
  nodes:
    ceos1:
      kind: arista_ceos
      image: ceos:4.32.0F
EOF
sudo containerlab deploy -t smoke.clab.yml

Wait ~1–2 min, then open the EOS CLI:

docker exec -it clab-ceos-smoke-ceos1 Cli

A ceos1> prompt + show version = cEOS runs on your Mac. 🎉 Tear it down:

sudo containerlab destroy -t smoke.clab.yml

7. Deploy the fabric

You're ready for the Course 2 lab. Drop in its ceos-evpn.clab.yml and:

cd ~/ceos-lab
sudo containerlab deploy -t ceos-evpn.clab.yml

The full 2-spine × 2-leaf + 2-host fabric takes ~5–8 min to boot under emulation. Always health-check before configuring (see the boot-race note below), then follow the lab guide.

---

Troubleshooting

Symptom	Cause	Fix
docker: command not found / can't connect	Docker context not set in the machine	It's usually automatic in OrbStack; re-enter with ssh orb, or check the OrbStack app is running
exec format error on container start	forgot --platform linux/amd64 on import	Re-import with the flag
Node boots, show interfaces Ethernet1 status shows type Unknown	cEOS boot-race — EOS scanned before containerlab wired the veths (worse under emulation)	containerlab destroy + deploy, then health-check until all nodes show connected / EbraTestPhyPort. Never docker restart a clab node — it destroys the veths (reload is also unavailable — it's a container)
Fabric slow to boot	4 emulated nodes booting at once	Normal — give it 5–8 min; watch with watch -n 5 'docker ps --filter name=clab-ceos-evpn --format "table {{.Names}}\t{{.Status}}"'
Can't reload a node	it's a container, not a box	Redeploy via containerlab; there's no in-box reboot

Health-check loop (run before configuring):

for n in spine1 spine2 leaf1 leaf2; do
  echo "== $n =="; docker exec clab-ceos-evpn-$n Cli -c "show interfaces Ethernet1 status"
done

Every node must show a real interface type (not Unknown) before you start.

---

Daily use

ssh orb                                              # enter the lab machine
cd ~/ceos-lab
sudo containerlab deploy  -t ceos-evpn.clab.yml      # bring the fabric up
sudo containerlab destroy -t ceos-evpn.clab.yml      # tear it down for the day
docker exec -it clab-ceos-evpn-leaf1 Cli             # jump into any node's CLI

Your Mac stays clean — the lab lives entirely inside the mylab machine, and you can destroy/deploy as often as you like. Next: the VXLAN-EVPN lab →.