Light Board Series: AHV Open vSwitch Networking – Part 4

We’re wrapping up our four part series on Nutanix AHV networking today with a look at the User VM Networking. Check out the Nutanix Connect Blog for full details.

We cover the difference between managed and unmanaged networks for VMs. VM networks can be rapidly created through the Prism GUI, the Acropolis CLI, or the REST API.

Light Board Series: AHV Open vSwitch Networking – Part 3

For part 3 in our series I want to tackle VLANs in AHV. I don’t actually have a light board video for this one 🙁

What I do have are some diagrams for you to look at!

Here’s the default VLAN configuration that we’d recommend:

acropolis_default_vlan

Here is a non-default configuration where a VLAN tag is added to the AHV host and the Controller Virtual Machine:
acropolis_custom_vlan

 

Learn more about VLANs in the Acropolis Hypervisor here on the official Nutanix NEXT Community Blog. Find complete details in the AHV Best Practice Guide.

 

Light Board Series: AHV Open vSwitch Networking – Part 1

I’m happy to announce the release of the first Light Board Videos I recorded with the Nutanix nu.school education team. These videos were a blast to record. The education team here at Nutanix is top notch and made my scribbles and rambling look and sound great! A video production team is an amazing asset to have sitting behind you in the office!

AHV provides an alternative to traditional hypervisors – and with that alternative comes a new virtual switch! This virtual switch bridges the VMs to the physical network.

To find more information about the video, including all of the rationale behind the decisions made – check out the Nutanix .NEXT Community blog I wrote describing AHV Host Networking.

Here’s the embedded first part of the video. I talk about Open vSwitch bridges and bonds, and how to connect the CVM and the User Virtual Machines to the 10gb or 1gb network interfaces. Follow the Nutanix .NEXT community blog, my site here, or the nu.school YouTube page to watch the rest of the series.

We’ll cover Load Balancing, Managed and Unmanaged VM networks, and more in the coming weeks!

Nutanix AHV Best Practices Guide

In my last blog post I talked about networking with Open vSwitch in the Nutanix Acropolis Hypervisor. Today I’m happy to announce the continuation of that initial post – the Nutanix Acropolis Hypervisor Best Practices Guide.

Nutanix Acropolis introduced the concept of AHV, based on the open source Linux KVM hypervisor. A new Nutanix node comes installed with AHV by default with no additional licensing required. It’s a full-featured virtualization solution that is ready to run VMs right out of the box. ESXi and Hyper-V are still great on Nutanix, but AHV should be seriously considered because it has a lot to offer, with all of KVMs rough edges rounded off.

Part of introducing a new hypervisor is describing all of the features, and then recommending some best practices for those features. In this blog post I wanted to give you a taste of the doc with some choice snippets to show you what this Best Practice Guide and AHV are all about.

Take a look at Magnus Andersson’s excellent blog post on terminology for some more detailed background on terms.

Acropolis Overview

Acropolis (one word) is the name of the overall project encompassing multiple hypervisors, the distributed storage fabric, and the app mobility fabric. The goal of the Acropolis project is to provide seamless invisible infrastructure whether your VMs exist in AWS, Hyper-V, ESXi, or the AHV. The sister project, Prism, provides the user interface to manage via GUI, CLI, or REST API.

Acropolis_Prism_Block_Diagram
AHV Overview

AHV is based on the open source KVM hypervisor, but is enhanced by all the other components of the Acropolis project. Conceptually, AHV has access to the Distributed Storage Fabric for storage, and the App Mobility Fabric powers the management plane for VM operations like scheduling, high availability, and live migration.

Acropolis Architecture CVM Scale

The same familiar Nutanix architecture exists, with a network of Controller Virtual Machines providing storage access to VMs. The CVM takes direct control of the underlying disks (SSD and HDD) with PCI passthrough, and exposes these disks to AHV via iSCSI (The blue dotted VM I/O line). The management layer is spread across all Nutanix nodes in the CVMs using the same web-scale principles of the storage layer. This means that by-default, a highly available VM management layer exists. No single point of failure anymore! No additional work to setup VM management redundancy – it just works that way.

AHV Networking Overview

Networking in AHV is provided by an Open vSwitch instance (OVS) running on each AHV host. The BPG doc has a comprehensive overview of the different components inside OVS and how they’re used. I’ll share a teaser diagram of the default network config after installation in a single AHV node.

acropolis_initial_install

AHV Networking Best Practices

Bridges, Bonds, and Ports – oh my. What you really want to know is “How do I plug this thing into my switches, setup my VLANs, and get the best possible load balancing. You’re in luck, because the Best Practice Guide covers the most common scenarios for creating different virtual switches and configuring load balancing.

Here’s a closer look at one possible networking configuration, where the 10gigabit adapters and 1gigabit adapters have been connected into separate OVS bridges. User VM2 has the ability to connect to multiple physically separate networks with this design to allow things like virtual firewalls.

acropolis_ovs_reco_2-10g

After separating network traffic, the next thing is load balancing. Here’s a look at another possible load balancing method called active-slb. Not only does the BPG provide the configuration for this, but also the rationale. Maybe fault tolerance is important to you. Maybe active-active configuration with LACP is important. The BPG will cover the config and the best way to achieve your goals.

For information on VLAN configuration, check out the Best Practices Guide.

Other AHV Best Practices

This BPG isn’t just networking specific. The standard features you expect from a hypervisor are all covered.

  • VM Deployment
    • Leverage the fantastic aCLI, GUI, or REST API to deploy or clone VMs.
  • VM Data Protection
    • Backup up VMs with local or remote snapshots.
  • VM High Availability
    • During physical host failure, ensure that VMs are started elsewhere in the cluster.
  • Live Migration
    • Move running VMs around in the cluster.
  • CPU, Memory, and Disk Configuration
    • Add the right resources to machines as needed.
  • Resource Oversubscription
    • Rules for fitting the most VMs onto a running cluster for max efficiency.

Take a look at the AHV Best Practice Guide for information on all of these features and more. With this BPG in hand you can be up and running with AHV in your datacenter and get the most out of all the new features Nutanix has added.

Networking Exploration in Nutanix AHV

Nutanix recently released the AHV hypervisor, which means I get a new piece of technology to learn! Before I started this blog post I had no idea how Open vSwitch worked or what KVM and QEMU were all about.

Since I come from a networking background originally, I drilled down into the Open vSwitch and KVM portion of the Nutanix solution. Here’s what I learned! Remember my disclaimer – I didn’t know anything about this before I started the blog. If I’ve got something a bit wrong feel free to comment and I’m happy to update or correct.

KVM Host Configuration

AHV is built on the Linux KVM hypervisor so I figured that’s a great place to start. I read the Nutanix Bible by Steve Poitras and saw this diagram on networking.

Networking diagram inside the Acropolis KVM Host
AHV OvS Networking

The CVM has two interfaces connecting to the hypervisor. One interface plugs into the Open vSwitch and the other goes to “internal”. I wasn’t sure what that meant. Looking through the hypervisor host config though I saw the following interfaces:

[root@DRM-3060-G4-1-1 ~]# ifconfig 
br0 Link encap:Ethernet HWaddr 0C:C4:7A:58:91:50 
  inet addr:10.59.31.77 Bcast:10.59.31.255 Mask:255.255.254.0
eth0 Link encap:Ethernet HWaddr 0C:C4:7A:3B:1C:8C 
eth1 Link encap:Ethernet HWaddr 0C:C4:7A:3B:1C:8D 
eth2 Link encap:Ethernet HWaddr 0C:C4:7A:58:91:50 
eth2.32 Link encap:Ethernet HWaddr 0C:C4:7A:58:91:50 
eth3 Link encap:Ethernet HWaddr 0C:C4:7A:58:91:51 
eth3.32 Link encap:Ethernet HWaddr 0C:C4:7A:58:91:51 
lo Link encap:Local Loopback 
virbr0 Link encap:Ethernet HWaddr 52:54:00:74:F9:B0 
  inet addr:192.168.5.1 Bcast:192.168.5.255 Mask:255.255.255.0
vnet0 Link encap:Ethernet HWaddr FE:54:00:9C:D8:CD 
vnet1 Link encap:Ethernet HWaddr FE:54:00:BE:99:B3

The next place I went was routing with netstat -r to see which interfaces were used for each next hop destination.

[root@DRM-3060-G4-1-1 ~]# netstat -r 
Kernel IP routing table
Destination  Gateway Genmask       Flags MSS Window irtt Iface
192.168.5.0  *       255.255.255.0 U     0 0 0           virbr0
10.59.30.0   *       255.255.254.0 U     0 0 0           br0
link-local   *       255.255.0.0   U     0 0 0           eth0
link-local   *       255.255.0.0   U     0 0 0           eth1
link-local   *       255.255.0.0   U     0 0 0           eth2
link-local   *       255.255.0.0   U     0 0 0           eth3
link-local   *       255.255.0.0   U     0 0 0           br0
default      10.59.30.1 0.0.0.0    UG    0 0 0           br0

I omitted a lot of text just to be concise here. We can see there are two interfaces with IPs, br0 and virbr0. Let’s start with virbr0, which is that internal interface. You can tell because it’s the 192.168 private IP used for CVM to hypervisor communication. I found that it was a local linux bridge, not an Open vSwitch controlled device:

[root@DRM-3060-G4-1-1 ~]# brctl show virbr0
bridge name bridge id         STP enabled interfaces
virbr0      8000.52540074f9b0 no          virbr0-nic
                                          vnet1

This bridge virbr0 has the vnet1 interface headed up to the internal adapter of the CVM – so THIS is where the CVM internal interface terminates.

That’s one side of the story – the next part is Open vSwitch

[root@DRM-3060-G4-1-1 ~]# ovs-vsctl show
be65c814-5d7c-46ab-bfb1-7b2bea19d954
 Bridge "br0"
  Port "tap345"
    tag: 32
    Interface "tap345"
  Port "vnet0"
    Interface "vnet0"
  Port "br0"
    Interface "br0"
      type: internal
  Port "bond0"
    Interface "eth2"
    Interface "eth3"
  Port "br0-dhcp"
    Interface "br0-dhcp"
      type: vxlan
      options: {key="1", remote_ip="10.59.30.82"}
  Port "br0-arp"
    Interface "br0-arp"
      type: vxlan
      options: {key="1", remote_ip="192.168.5.2"}
ovs_version: "2.1.3"

OvS has a vSwitch called br0. The CVM vnet0 is a port on this bridge, and so is bond0 (the combination of the 10GbE interfaces). We also see this special “type:internal” interface – this one with the IP address assigned to it. This is the external facing IP of the AHV / KVM hypervisor host.

[root@DRM-3060-G4-1-1 network-scripts]# cat ifcfg-br0 
DEVICE=br0
DEVICE_TYPE=ovs
TYPE=OVSIntPort
NM_CONTROLLED=no
ONBOOT=yes
BOOTPROTO=none
IPADDR=10.59.31.77
NETMASK=255.255.254.0
GATEWAY=10.59.30.1
OVSREQUIRES="eth3 eth2 eth1 eth0"

In addition to the CVM, external, and internal interfaces we see a tap345 interface tagged in VLAN 32. This matches the tagged interfaces from our “ifconfig -a” command above: eth2.32 and eth3.32. It’ll be used for a VM that has a network interface in VLAN 32.

Finally – we come to the IP Address Management (IPAM) interfaces, br0-arp, and br-dhcp. Steve mentions VXLAN and here’s where we see those concepts. The OvS can either intercept and respond to DHCP traffic, or just let it through. If we allow OvS to intercept the traffic this means Acropolis and Prism now become the point of control for giving out IP addresses to VMs that boot up. Very cool!

Now let’s take a look at the config parameters passed to the running CVM. Right now this box has ONLY the CVM running on it so only one instance of qemu-kvm running.

[root@DRM-3060-G4-1-1 ~]# ps -ef | grep qemu
qemu 9250 1 61 Jun26 ? 1-21:13:21 /usr/libexec/qemu-kvm -name NTNX-DRM-3060-G4-1-1-CVM 
-S -enable-fips -machine pc-i440fx-rhel7.0.0,accel=kvm,usb=off,mem-merge=off -cpu host,+kvm_pv_eoi -m 24576 -realtime mlock=on -smp 8,sockets=8,cores=1,threads=1 
-uuid 1323cbbc-a20d-d66a-563e-ca7a8609cb73 -no-user-config -nodefaults 
-chardev socket,id=charmonitor,path=/var/lib/libvirt/qemu/NTNX-DRM-3060-G4-1-1-CVM.monitor,server,nowait 
-mon chardev=charmonitor,id=monitor,mode=control 
-rtc base=utc -no-shutdown -boot menu=off,strict=on 
-kernel /var/lib/libvirt/NTNX-CVM/bzImage -initrd /var/lib/libvirt/NTNX-CVM/initrd 
-append init=/svmboot quiet console=ttyS0,115200n8 
-device piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 
-netdev tap,fd=20,id=hostnet0,vhost=on,vhostfd=21 
-device virtio-net-pci,netdev=hostnet0,id=net0,mac=52:54:00:9c:d8:cd,bus=pci.0,addr=0x3 
-netdev tap,fd=22,id=hostnet1,vhost=on,vhostfd=23 
-device virtio-net-pci,netdev=hostnet1,id=net1,mac=52:54:00:be:99:b3,bus=pci.0,addr=0x4 
-chardev file,id=charserial0,path=/tmp/NTNX.serial.out.0 
-device isa-serial,chardev=charserial0,id=serial0 
-vnc 127.0.0.1:0 -vga cirrus 
-device pci-assign,configfd=24,host=01:00.0,id=hostdev0,bus=pci.0,addr=0x5,rombar=0 
-device virtio-balloon-pci,id=balloon0,bus=pci.0,addr=0x7 -msg timestamp=on

Maybe a better way to look at the CVM details is via the XML configuration:

[root@DRM-3060-G4-1-1 ~]# cat /etc/libvirt/qemu/NTNX-DRM-3060-G4-1-1-CVM.xml
<domain type='kvm'>
 <name>NTNX-DRM-3060-G4-1-1-CVM</name>
 <uuid>1323cbbc-a20d-d66a-563e-ca7a8609cb73</uuid>
 <memory unit='KiB'>25165824</memory>
 <currentMemory unit='KiB'>25165824</currentMemory>
 <memoryBacking>
 <nosharepages/>
 <locked/>
 </memoryBacking>
 <vcpu placement='static'>8</vcpu>
 <os>
 <type arch='x86_64' machine='pc-i440fx-rhel7.0.0'>hvm</type>
 <kernel>/var/lib/libvirt/NTNX-CVM/bzImage</kernel>
 <initrd>/var/lib/libvirt/NTNX-CVM/initrd</initrd>
 <cmdline>init=/svmboot quiet console=ttyS0,115200n8</cmdline>
 <boot dev='hd'/>
 <bootmenu enable='no'/>
 </os>
 <features>
 <acpi/>
 <apic eoi='on'/>
 <pae/>
 </features>
 <cpu mode='host-passthrough'>
 </cpu>
 <clock offset='utc'/>
 <on_poweroff>destroy</on_poweroff>
 <on_reboot>restart</on_reboot>
 <on_crash>restart</on_crash>
 <devices>
 <emulator>/usr/libexec/qemu-kvm</emulator>
 <controller type='usb' index='0'>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x2'/>
 </controller>
 <controller type='ide' index='0'>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x01' function='0x1'/>
 </controller>
 <controller type='pci' index='0' model='pci-root'/>

##Here is the CVM interface being linked to the br0 OvS interface
 <interface type='bridge'>
 <mac address='52:54:00:9c:d8:cd'/>
 <source bridge='br0'/>
 <virtualport type='openvswitch'>
 <parameters interfaceid='7a0a4887-b8cd-4f02-960d-cca5c1ca73cc'/>
 </virtualport>
 <model type='virtio'/>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/>
 </interface>

##Here is the CVM interface going to the local link bridge.
 <interface type='network'>
 <mac address='52:54:00:be:99:b3'/>
 <source network='NTNX-Local-Network'/>
 <model type='virtio'/>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x04' function='0x0'/>
 </interface>

 <serial type='file'>
 <source path='/tmp/NTNX.serial.out.0'/>
 <target port='0'/>
 </serial>
 <console type='file'>
 <source path='/tmp/NTNX.serial.out.0'/>
 <target type='serial' port='0'/>
 </console>
 <input type='mouse' bus='ps2'/>
 <input type='keyboard' bus='ps2'/>
 <graphics type='vnc' port='-1' autoport='yes' listen='127.0.0.1'>
 <listen type='address' address='127.0.0.1'/>
 </graphics>
 <video>
 <model type='cirrus' vram='16384' heads='1'/>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/>
 </video>
 <hostdev mode='subsystem' type='pci' managed='yes'>
 <source>
 <address domain='0x0000' bus='0x01' slot='0x00' function='0x0'/>
 </source>
 <rom bar='off'/>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x05' function='0x0'/>
 </hostdev>
 <memballoon model='virtio'>
 <address type='pci' domain='0x0000' bus='0x00' slot='0x07' function='0x0'/>
 </memballoon>
 </devices>
</domain>

We saw a new reference in that last command, NTNX-Local-Network. If we look at virsh for information about defined networks we see the following:

[root@DRM-3060-G4-1-1 ~]# virsh net-list --all 
 Name               State  Autostart Persistent
----------------------------------------------------------
 NTNX-Local-Network active yes       yes
 VM-Network         active yes       yes

If we look in the /root/ partition there are definitions for these:

[root@DRM-3060-G4-1-1 ~]# cat net-NTNX-Local-Network.xml 
<network connections='1'>
 <name>NTNX-Local-Network</name>
 <bridge name='virbr0' stp='off' delay='0' />
 <ip address='192.168.5.1' netmask='255.255.255.0'>
 </ip>
</network>

[root@DRM-3060-G4-1-1 ~]# cat net-VM-Network.xml 
<network connections='1'>
 <name>VM-Network</name>
 <forward mode='bridge'/>
 <bridge name='br0' />
 <virtualport type='openvswitch'/>
 <portgroup name='VM-Network' default='yes'>
 </portgroup>
</network>

These two pieces of information tie everything together neatly for us. The internal network given to the CVM is the linux virbr0 device. The external network given to the CVM is OvS br0.

Now I think I finally understand that image presented at the beginning!

CVM Guest Configuration

Since we understand the KVM/AHV host configuration lets take a look in the CVM guest. This should be a little easier.

nutanix@NTNX-15SM60140129-A-CVM:10.59.30.77:~$ netstat -r
Kernel IP routing table
Destination Gateway Genmask         Flags MSS Window irtt Iface
192.168.5.0 *       255.255.255.128 U     0 0 0           eth1
192.168.5.0 *       255.255.255.0   U     0 0 0           eth1
10.59.30.0  *       255.255.254.0   U     0 0 0           eth0
link-local  *       255.255.0.0     U     0 0 0           eth0
link-local  *       255.255.0.0     U     0 0 0           eth1
default     10.59.30.1 0.0.0.0      UG    0 0 0           eth0

The routing table shows the internal and external networks, and just two network adapters. The eth1 adapter has a subinterface. This one is eth1:1 as opposed to .1. Not entirely sure what that one means – but I’ll keep it in mind in case I come across something later on.

nutanix@NTNX-15SM60140129-A-CVM:10.59.30.77:~$ ifconfig -a
eth0 Link encap:Ethernet HWaddr 52:54:00:9C:D8:CD 
  inet addr:10.59.30.77 Bcast:10.59.31.255 Mask:255.255.254.0

eth1 Link encap:Ethernet HWaddr 52:54:00:BE:99:B3 
  inet addr:192.168.5.2 Bcast:192.168.5.127 Mask:255.255.255.128

eth1:1 Link encap:Ethernet HWaddr 52:54:00:BE:99:B3 
  inet addr:192.168.5.254 Bcast:192.168.5.255 Mask:255.255.255.0
  UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1

lo Link encap:Local Loopback 
  inet addr:127.0.0.1 Mask:255.0.0.0

That’s it – just two simple interfaces in the CVM. One for internal traffic to the hypervisor directly, another for receiving any external requests from remote CVMs, the management APIs, and all of the other magic that the CVM performs!

This concludes our walkthrough of networking inside a Nutanix AHV machine. I hope you learned as much as I did going through these items! Please comment or reach out to me directly if you have any questions.