Building a Virtual Lab with Arista vEOS and VirtualBox
It’s no secret that I love Arista switches. When I wrote Arista Warrior, I was lucky enough to have a loaner switch from Arista in my home lab, but sadly they made me give it back. Since Arista is a relative newcomer to the world of Networking, there isn’t a pile of used Arista gear on eBay, so I can’t build a killer lab at home without spending thousands of dollars. As much as I love Arista switches, I’d rather spend my spare cash on
guitars my wife and kids.
Understanding the plight of cash-strapped networking guys the world over, Arista has released a virtual-machine-ready version of their fabulous switch operating system, EOS. Currently this is only available to existing Arista customers, so see your Arista sales rep to get a copy. Please don’t ask me for a copy, since I will not send you a copy no matter how much you beg. Arista has hinted that they may release this into the general population, in which case I may build a Virtual Box appliance to share. Until then, you’ll need to read on and build it yourself.
I’ve built this lab using my Macbook Pro that has an Intel i7 processor with 16GB of RAM. The OS in use is Mountain Lion (OSX v10.7.5). I’m using VirtualBox for Mac v4.2.1 (r80871), and vEOS v 4.10.2. I’ve configured each VM to have one CPU, and 1GB of RAM. Real Arista switches ship with at least 4GB of RAM, but this is a lab after all, so 1GB should be fine. You can further tune them down to 512MB each, and they’ll run, but they’ll have little (if any) free memory. If you’ve got the RAM, I’d stick to 1GB for each VM.
Virtual machines are possible using a variety of tools, so why did I choose VirtualBox? Mainly because it’s free, and this way anyone with access to the vEOS files would be able to follow along. Also because it’s available on many host operating systems. Did I mention that it’s free? I love free things, especially when they’re useful!
Using the lab as shown requires that you have a clue about linux, because I will not go into detail about installing it. You can build it without the Linux server if you’d like, and the switches will still work. The Linux server adds useful functionality for things like uploading code, learning ZTP, and so-on.
The Network We’ll Build
We’ll be building the network shown to the right. I originally started with a different concept, but quickly discovered that VirtualBox only supports four interfaces per VM. I then discovered that the first virtual network interface I created always ended up being the Management1 interface in each switch, which further limited me to three network interfaces in each switch.
The virtual Arista switches have 1Gbps Management interfaces, while all the other interfaces are 10Gbps. Whether or not they can actually deliver that speed is unimportant to me – I just wanted to make sure that all of the interfaces used for inter-switch links were the same. Additionally, I had another driver for keeping the management interfaces separate.
If you really want to learn how to use a networking device, you’ll want to do things like TFTP images from a server. Additionally, EOS includes cool features like Zero Touch Provisioning (ZTP) that make use of options in DHCP. To that end, I felt it to be beneficial to include a fourth VM running Ubuntu Linux.
I’ve structured the lab in such a way that the links between each switch act like physical cables (Separate internal Networks in VirtualBox), but the network connecting the management interfaces to the Ubuntu Linux server are on a common shared (host-only) network . Lastly, I’ve configured the Linux box to have access to the outside world by having one of its network interfaces configured as a NAT interface in virtual box. Given this design, code, extensions, and anything else you’d like can be downloaded from the Internet to the Linux box, then the virtual switches can retrieve it from there. The virtual switches cannot get outside of the virtual lab, so you should be able to abuse them in any way imaginable without fear of corrupting the networks outside of your computer.
Enough Blather, Let’s get to It!
The first thing you’ll need to do is download and install Oracle’s Virtual box. You can get it for your OS here: https://www.virtualbox.org. You’ll also need the Arista vEOS files, which you can get from Arista if you’re an existing customer.
There are actually two Arista files that are needed: the virtual hard drive, which is in the .vmdk cross-platform format (works with any VM software that supports it), and the Aboot ISO file, which is specific to each VM platform. The source I got from Arista included files for VirtualBox, VMWare, HyperV, and QEMU. The files I’ll be using are as follows. Of course the file names may change depending on the EOS and Aboot releases:Aboot-veos-virtualbox-2.0.6.iso EOS-4.10.2-veos.vmdk
Find the location where VirtualBox will place the VMs, create a folder there, and put both of these files in that folder. On my Mac, I’ve placed them into Users/GAD/VirtualBox VMs/AristaFiles/. The reason for this will make sense in a minute.
Build a Base VM
Because of the way VirtualBox seems to work, and because of my obsessive nature, we’ll build a base VM, then clone it three times. This will create three VMs, each with its own folder, and each having its own virtual disk. This will be of use should we ever decide to upgrade the image in the future, as it will allow each switch to run separate versions.
With VirtualBox installed, and the Arista files in the right location, we can now create the base VM. In VirtualBox, click on the New button. Name the new VM whatever you’d like (I’ve used vEOS-Base). For Type, choose Linux, and for Version, choose Other Linux. Though EOS is based on Fedora, I’ve had the best luck using Linux/Other Linux. Feel free to experiment, but remember that any examples I post will use this setting. Hit Continue when you’re ready.
You should now choose how much memory to allocate for the VM. I’ve used 1024MB on all VMs. I’ve successfully created them with as little as 512MB, but they had little to no free memory, so I’d recommend 1024MB or more. Remember that whatever you choose will be replicated three times, so in my case with a 1024MB VM, I will end up consuming 3GB of RAM with the Arista vEOS VMs alone. Once you’ve specified the memory size, click continue, which should bring you to the hard drive screen.
There should be three options: Do not add a virtual HD, Create a virtual HD now, and Use an existing virtual HD. We want the last one, Use an existing virtual HD, so choose that. We’ll now need to specify the location of the virtual hard drive. Click the little button next to the pull-down list, and specify the .vmdk file from the folder we created above. In my case, the file name is EOS-4.10.2-veos.vmdk located in the Users/GAD/VirtualBox VMs/AristaFiles/ folder. Once that’s chosen, click the Create button, and you’re first Virtual Arista switch will be created. Don’t get too excited, though, because it won’t work yet.
With the newly created VM selected in the left pane of VirtualBox, click the Settings button. We’ll need to tweak some settings to make vEOS boot. Within the settings window, choose the Systems button, and then within that screen, choose the Processor tab. In this screen there should be a section called Extended Features with an entry that reads Enable PAE/NX. Check this box to make sure it is enabled. Though this doesn’t seem to matter on my Mac (it works either way), when I tried this on Windows, the VM would not boot unless this box was checked. When done, click the Storage button on the top to being us to the Storage Settings window.
vEOS is very particular about how its storage is configured. The virtual hard drive we already added should be where it belongs, but we need to add the Aboot image for the VM to boot. Aboot is the (very cool) boot loader used on Arista switches, and it needs to be installed as a CD, but it must be configured on the IDE Primary Slave on the first (usually only) IDE controller. If a SCSI controller gets created, it must be deleted or vEOS will not load.
To configure the Aboot ISO image, first make sure that there is only one IDE controller in the left pane. In the right pane, click on the pulldown menu and choose IDE Primary Slave. Click the little CD icon next to the pulldown, and choose Choose a Virtual CD/DVD Disk File from the list. Navigate to the location we created above, and choose the Aboot image for VirtualBox. In my examples, it is named Aboot-veos-virtualbox-2.0.6.iso and is located in the Users/GAD/VirtualBox VMs/AristaFiles/ folder. Double-click the file, or click the Open button when done, which should return you to the Storage settings window. This window should now look something like the one to the right. Click OK to save the settings.
At this point, you should have a bootable vEOS virtual machine. Go ahead and click the start button to load the VM. If it doesn’t boot, check all of your settings to make sure they’re correct. If this VM doesn’t’ boot, we can’t continue, so make sure it works before moving on to the next section. Sometimes these VMs take a while to boot, especially the first time, so as long as you’re seeing [ OK ] messages, you’re probably doing fine. If you hang at Starting New Kernal for more than a minute or so, there’s probably something wrong – check your images and your IDE configuration. If you’re waiting at Completing EOS Initialization for more than a few minutes, there may be something wrong, but it really depends on the resources available to your machine. Be patient, and try to remember that these are very powerful switches being shoehorned into little VMs.
Once the VM boots, you can log in using the admin username. This switch will have only one interface – Management1, which you can configure if you really want to, but for now let’s keep it simple since this isn’t really our end-state anyway. Feel free to power it off and we’ll move on to the cloning process.
Cloning the VMs.
With the base vEOS VM built, we can now clone it to make our actual switch VMs. To do so, right click on the vEOS-Base VM in the left pane, and choose Clone. This will prompt you for a new VM name, and the option to Reinitialize the MAC address of all network cards. Make sure this box is checked, because it would be pretty useless to have three switches that all have the same MAC address. When ready, click the Continue button. This will bring you to a page asking what type of clone you want (Full clone or Linked Clone). Choose Full Clone, which should be the default. When that’s done, click the Clone button.
Do this step three times, naming the new VMs vEOS-1, vEOS-2 and vEOS-3. At this point, you should have four VMs, including the original vEOS-Base VM. The difference is that each of the clones has its own .vmdk file, while the original points to the .vmdk file in the Arista Files folder. The Aboot image points to the Arista Files folder in each VM, but this is a read-only ISO, so that’s OK. If you’d like to change this, copy the Aboot image to each of the VM folders and then change where the image links in the Settings/Storage page for each VM.
At this point, you can delete the vEOS-Base image, since we’ll no longer be using it. To do so, right-click on the VM in the left pane of VirtualBox, and choose Remove. You will be given the options of Remove only and Delete all files. Do not delete all files! If you do, you will lose the original .vmdk file. Hopefully you have it saved somewhere else, but be warned that it will be gone if you delete the files. When that’s done, there should be only three VMs: vEOS-1, vEOS-2, and vEOS-3.
At this point, go ahead and create the Ubuntu Linux VM. You can use any Linux you want, actually, and I won’t go into the details of installing that because it should be pretty straightforward. I recommend that you download and install the server version (http://www.ubuntu.com/download/server) since it’s smaller than the desktop version. After that’s done, it’s time to build the networks.
Building the Virtual Networks
VirtualBox supports multiple network types for the VMs. There are three that we’ll be using: Internal Networks, Host-only networks, and NAT. An Internal network is one that exists only within the realm of VirtualBox, and is only included on interfaces configured to use it. By creating multiple internal networks, we will simulate physical cables connecting the switches. In other words, each internal network will simulate a physical cable. I’ve named these internal networks according to the devices using them. For example the link between vEOS-1 and vEOS-3 is named NET-13. The link between vEOS-2 and vEOS-3 is named NET-23. There are two links between vEOS-1 and vEOS-2, so they are named NET-121 and NET-122. These networks are all shown on the drawing to the right.
The management network is shared, so each switch can see every other switch, as well as the Ubuntu Linux VM on this network. This is done because the Management interfaces are routed, and so that each switch can get to the Ubuntu Linux server in the same way.
Lastly, the Ubuntu Linux server also has a NAT interface, which allows it to connect to the Internet through the host operating system (in my case, Mac OSX). Let’s go ahead and build those networks now.
With all of the VMs shut down and powered off, click on vEOS-1 and then click on Settings. Within the Settings window, click on the Network button. There should be four network adapters, named Adapter 1, 2, 3, and 4. Adapter 1 will be the Management1 interface on the vEOS devices. Make sure that Enable Network Adapter is checked. On the Attached to pulldown menu, choose Host-only Adapter. On a Mac, the Name will default to vboxnet0. On Windows, this will read Host-only network adapter. It needs to match Adapter1 in every vEOS VM. Adapter2 in the Ubuntu Linux VM should be configured this way as well.
On the same page, choose the Advanced menu by clicking on the little triangle. This will open more options, a couple of which are required for our virtual switches to work. First is the Adapter Type. More than one of these will work, but the one that works best in my experience is PCnet-FAST III. Originally I had recommended the Intel PRO/1000 T interface, but I’ve since learned that this type may strip VLAN tags which will cause your trunks (including MLAG peer links) to break. I know this says that it’s a 1Gbps interface, but the vEOS VMs will report 10Gbps, so don’t worry about that.
The next option is Promiscuous Mode, and this must be set to Allow VMs or Allow All. I have all of mine set to Allow VMs. Lastly, The checkbox for Cable Connected should be checked.
Next, configure Adapter2 on the Ubuntu Linux VM to be attached to NAT. There are no other options. When this is done, the VM configuration for the Ubuntu Linux VM is complete (though we’ll still need to configure the OS with IP addresses later on).
Now we need to configure all of the Internal Networks within the vEOS VMs. I’ll show one example, then list how they should be configured in the interest of saving space. For our example, we’ll configure vEOS-1, Adapter2.
Look at the network drawing at the beginning of this section. This drawing shows where each adapter is, what network it’s configured for, and how the VMs are interconnected. Looking at vEOS-1, we can see that Adapter2 translates to Eth1, which connects to vEOS-2 on Eth1, and the physical cable is represented as NET-121. To configure this, go into Settings, Network for vEOS-1, and choose Adapter2. Make sure the interface is enabled, choose Internal Network from the Attached to pulldown, then enter NET-121 in the Name field. Remember to click the advanced section and set the adapter type to PCnet-FAST III, Promiscuous Mode to Allow VMs and check checkbox for Cable Connected.
Configure the remaining network adapters on each VM as shown in the spreadsheet on the right. When this is done, the lab should be bootable, and every device should be able to see every other device according to the original network diagram. To prove if this is working, use the show lldp neighbor command from within vEOS. Here’s an example from vEOS-1 which is configured correctly. It’s OK to see all switches on the Ma1 interface since that network is shared. Since these are routed interfaces, there should be no interference with L2 protocols like spanning-tree.
vEOS-1#sho lldp nei Last table change time : 0:01:41 ago Number of table inserts : 5 Number of table deletes : 0 Number of table drops : 0 Number of table age-outs : 0 Port Neighbor Device ID Neighbor Port ID TTL Et1 vEOS-2 Ethernet1 120 Et2 vEOS-2 Ethernet2 120 Et3 vEOS-3 Ethernet1 120 Ma1 vEOS-2 Management1 120 Ma1 vEOS-3 Management1 120
Here’s the output from the show spanning-tree command on vEOS-1:
vEOS-1#sho spanning-tree MST0 Spanning tree enabled protocol mstp Root ID Priority 32768 Address 0800.27e4.d8fb This bridge is the root Bridge ID Priority 32768 (priority 32768 sys-id-ext 0) Address 0800.27e4.d8fb Hello Time 2.000 sec Max Age 20 sec Forward Delay 15 sec Interface Role State Cost Prio.Nbr Type ---------------- ---------- ---------- --------- -------- -------------------- Et1 designated forwarding 2000 128.1 P2p Et2 designated forwarding 2000 128.2 P2p Et3 designated forwarding 2000 128.3 P2p
The Final Touches
The last thing to do to make this lab really useful is to assign IP addresses to the management interfaces (see the spreadsheet in the previous section for how I have mine configured), and the Ubuntu Linux server. To set the Ubuntu Linux server, edit the /etc/network/interfaces file with the sudo vi /etc/network/interfaces command (You can use vi right?) and make sure the following is included in the file:
auto eth0 iface eth0 inet dhcp auto eth1 iface eth1 inet static address 10.10.10.10 netmask 255.255.255.0 network 10.10.10.0 broadcast 10.10.10.255
Save the file and issue the sudo /etc/init.d/networking restart command. At this point, you should be able to ping every VM from every other VM using the management network.
arista@Ubuntu:/etc/network$ ping -c 5 10.10.10.11 PING 10.10.10.11 (10.10.10.11) 56(84) bytes of data. 64 bytes from 10.10.10.11: icmp_req=1 ttl=64 time=0.287 ms 64 bytes from 10.10.10.11: icmp_req=2 ttl=64 time=3.15 ms 64 bytes from 10.10.10.11: icmp_req=3 ttl=64 time=0.624 ms 64 bytes from 10.10.10.11: icmp_req=4 ttl=64 time=0.524 ms 64 bytes from 10.10.10.11: icmp_req=5 ttl=64 time=0.416 ms --- 10.10.10.11 ping statistics --- 5 packets transmitted, 5 received, 0% packet loss, time 4013ms rtt min/avg/max/mdev = 0.287/1.001/3.155/1.082 ms
Finally, we’ll add one more IP address so that the host machine can also get to the management network. You don’t really need to do this, and it can be accomplished in other ways, but I like this because it works and I’m lazy. Open preferences in VitualBox (for the application, not for a VM), click on the Network icon, and you should be presented with a list of Host-only networks. There should be only one unless you have other VMs built. Double click on the network (vboxnet0 on the Mac), and you should get a new window with two tabs: Adapter and DHCP server. In the adapter tab, assign an IP address of 10.10.10.1 with an IPv4 mask of 255.255.255.0.
Next, click on the DHCP Server tab, and uncheck Enable Server. The reason for this is that we don’t want VirtualBox to dole out IP address – we want the Ubuntu Linux VM to do this. This is because if and when we play around with Zero Touch Provisioning, we’ll configure advanced settings in Linux that aren’t available in Virtual Box.
Here’s one more cool benefit of the way we’ve built this lab. Add the user arista with the password arista to every device. On the vEOS machines, use the following exec mode command:
vEOS-1#conf t vEOS-1(config)#username arista privilege 15 secret arista vEOS-1(config)#wri
On the Ubuntu Linux machine, use this command:
sudo useradd -b /home/arista -m -p arista arista
Once the usernames have been added to each device, you can use your favorite SSH client to connect to each VM, including the Ubuntu Linux machine. This is cool because though you can do everything from the console of each VM, using the console sucks compared to the power of a full-featured SSH client like Secure-CRT. At this point, the Lab is complete, and you can abuse it to your heart’s content. Congratulations!
Sometimes, no matter how careful we are, things just don’t work right. Here are some common errors and problems I’ve seen while building and endlessly rebuilding this lab.
FATAL: No bootable medium found! System Halted
This error means that the VM cannot find the Aboot ISO file, or the file is corrupted.
VM hangs at Starting new kernel forever
This is indicative of the VM being unable to find the virtual hard disk (.vmdk) file. Check your storage settings and make sure the .vmdk for that VM exists.
VM is slooooow
If it’s working, but it’s really slow, try giving the VM more memory. If you’ve only got 4GB of RAM in your machine, this lab will probably not run very well, so you might just have to live with it. Ideally, this lab needs a minimum of 4 CPUs and 4GB of RAM.
I hope this helps people who need an Arista lab. I know it’s helped me quite a bit. I might even be inclined to write up some lessons using this lab, so stay tuned and enjoy my other ramblings while you wait.