If you rather download this article as a PDF, click here.Jose De la Rosa
Dell Linux Engineering
This article describes how to quickly setup and manage a virtualized environment with KVM (Kernel-based Virtual Machine) in Red Hat Enterprise Linux 6 on a Dell PowerEdge server. This is not an in-depth discussion of virtualization or KVM, but rather an easy-to-follow step-by-step description of how to install and manage Virtual Machines (VMs) on a physical server. Perhaps the biggest advantage of virtualization is cost savings. Even with the simplest deployment you will save hardware ($$) and management (time) costs. For example in my lab I used to have 3 separate physical servers that ran all my services and apps, now I have only one physical server with 4 VMs and plenty or resources (CPUs, memory & disk space) to spare. In this article I share the exact same steps I followed to install and manage the virtual environment in my lab.Note #1: I don’t use any Windows VMs in my lab, so I will only cover Linux VMs. The steps for installing Windows VMs are basically the same, with some additional configuration steps that I will cover in detail in a future article.Note #2: I don’t use GUIs much, so I will only cover using a shell terminal. Come on folks this is Linux, GUIs are more Windows users.
A line or two about KVM and then we’ll move on to the fun stuff. KVM is a Linux kernel module that allows a user space program access to the hardware virtualization features of Intel and AMD processors. With the KVM kernel module, VMs run as ordinary user-space processes.KVM uses QEMU for I/O hardware emulation. QEMU is a user-space emulator that can emulate a variety of guest processors on host processors with decent performance. Using the KVM kernel module allows it to approach native speeds. KVM is managed via the libvirt API and tools. Some libvirt tools used in this article include virsh, virt-install and virt-clone.
Let’s agree on the terms & conventions I use here:
You don’t need a top-of-the-line PowerEdge server, a 2-socket server (with 4-6 cores each), 8-16 GB of memory and 100 GB disk space is sufficient for 4 to 8 VMs, but of course it will depend on your needs. In my lab I have a PowerEdge T610 with 2 Intel Xeon 4-core 5520 processors, 16 GB of memory and about 160GB disk space and I can easily run 8-10 RHEL VMs without a sweat. Most of my VMs use 1 GB of memory, 1 virtual CPU (VCPU) and around 15 GB of disk space.One awesome feature in KVM is the concept of overcommitting, which allows you to allocate more virtualized CPUs and memory than available resources on the host. The VMs then only use what they need, allowing other VMs to use unused resources. So when you are planning your VM resources, remember that you can allocate more resources than you actually have, but of course be mindful that the total amount of resources actually being used by all VMs can’t exceed the total amount of resources available on your host.
Verify that Virtualization Technology (VT) is enabled in your server’s BIOS. Reboot your server and press F2 during POST to go into the BIOS, the select “Processors Settings”, and verify “Virtualization Technology” is “Enabled”. Save and Exit and let the server boot up.Another thing to check once your server boots up is whether your processors support VT. This is apparently not a requirement but it will help a lot with performance, you might be better off with processors that do support VT. Check for these CPU extensions:# grep -E 'svm|vmx' /proc/cpuinfo- vmx is for Intel processors- svm is for AMD processors
Verify you have enough disk space on the host. KVM creates image files (which look just like regular files in the file system) for each VM and assumes you are going to place them in directory /var/lib/libvirt/images. Since the default RHEL installation doesn’t create a separate partition for /var (it places everything under ‘/’) I don’t like to use /var/lib/libvirt/images. Instead I create a separate partition for my VM images, I like /vm-images. KVM supports several types of VM image formats, which determine the amount of actual disk space each VM uses on the host. In this article, we will only create VMs with raw file formats, which use the exact amount of disk space you specify. So for example if you specify that a VM will have 10 GB of disk space, the VM install tool will create a file image of exactly 10 GB on the host, regardless whether the VM uses all 10 GB or not.Best practice here is to allocate more than enough disk space on the host to safely fit all your VMs. For example, if you want to create 4 VMs with 20GB storage each, be sure you have at least 85-90 GB space available on your host. Disk space is cheap nowadays, so hopefully this will not be an issue.
There are several RPMs to install that are not part of the base RHEL 6 installation. Assuming that you have a yum repository, install the following: # yum install kvm python-virtinst libvirt libvirt-python virt-manager virt-viewer libguestfs-toolsVerify the following kernel modules are loaded, and if not load manually:kvmkvm_intel (only on Intel-based systems)
By default, the VMs will only have network access to other VMs on the same server (and to the host itself) via private network 192.168.122.0. If you want the VMs to have access to your LAN, then you must create a network bridge on the host. Follow these steps to create a network bridge: 1. Turn off NetworkManager (the network bridge does not like it) and use the 'network' service instead. Be sure to set NM_CONTROLLED=NO in your network controller configuration file (presumably /etc/sysconfig/network-scripts/ifcfg-em1):
# chkconfig NetworkManager off# chkconfig network on# service NetworkManager stop# service network start2. Add to your network controller configuration file the following line:BRIDGE=br03. Create /etc/sysconfig/network-scripts/ifcfg-br0 and add:DEVICE="br0"# BOOTPROTO is your preference. it can be "dhcp" or "static".# If "static", be sure to specify the IP address, netmask and gateway. BOOTPROTO="dhcp"IPV6INIT="yes"IPV6_AUTOCONF="yes"NM_CONTROLLED="no"ONBOOT="yes"TYPE="Bridge"DELAY="0"4. Create a FORWARD firewall rule for the bridge br0 so that VM network traffic can be forwarded through it. This rule works for all bridge devices:# iptables -A FORWARD -m physdev --physdev-is-bridged -j ACCEPT# service iptables save5. Enable forwarding. Edit /etc/sysctl.conf:inet.ipv4.ip_forward = 1And read the file:# sysctl -p /etc/sysctl.conf6. Restart the ‘network’ service so that the network bridge you just created can get an IP address:# service network restart
If you are using SELinux in Enforcing mode, then there are some things to consider. The most common problem is when you use a directory for your VM images other than the default directory /var/lib/libvirt/images, since you will need to change the security context for the directory you selected. For example, let’s say you select /vm-images to place your VM images:1. Create the directory:# mkdir /vm-images2. Install the policycoreutils-python package (which contains the semanage SELinux utility):# yum -y install policycoreutils-python3. Set the security context for the directory and everything under it:# semanage fcontext --add -t virt_image_t '/vm-images(/.*)?'Verify it:# semanage fcontext -l | grep virt_image_t4. Restore the security context. This will effectively change the context to virt_image_t:# restorecon -R -v /vm-imagesVerify that the context was changed:# ls -aZ /vm-imagesdrwxr-xr-x. root root system_u:object_r:virt_image_t:s0 .dr-xr-xr-x. root root system_u:object_r:root_t:s0 ..5. If you are going to export the directory /vm-images as a samba or NFS share, there are SELinux Booleans that need to be set as well:# setsebool -P virt_use_samba 1# setsebool -P virt_use_nfs 1
Finally, you need to have an OS installation source ready. You can either have a RHEL iso on your host or a remote installation source which you can use via http, ftp or nfs.
Installation of VMs using the virt-install tool is very straight-forward. This tool can run in interactive or non-interactive mode. Let’s use virt-install in non-interactive mode to create a RHEL 6.1 x64 VM named vm1 with one virtual CPU, 1 GB memory and 10 GB of disk space:# virt-install \--network bridge:br0 \--name vm1 \--ram=1024 \--vcpus=1 \--disk path=/vm-images/vm1.img,size=10 \--graphics none \ --location=http://my.server.com/pub/rhel6.1/install-x86_64/ \--extra-args="console=tty0 console=ttyS0,115200"
If you created a network bridge (as specified in Chapter I, steps 6-10) and want to use it for full inbound and outbound connectivity, then you must specify it.
No big mystery here, this is the name of the VM
This is the amount of memory in the VM in MBs
You guessed it, this is the number of virtual CPUs
This is the image file for the VM, the size is specified in GBs.
This tells the installer not to launch a VNC window to access the VM’s main console. Instead, it will use a text console on the VM’s serial port. If you rather use an X window with graphics to install the OS on the VM, omit this parameter.
This is the location of the RHEL 6.1 x64 installation directory, which of course will be different for you. If you don’t have a remote installation location for the OS, you can install from an iso instead. Instead of using the location parameter, use the cdrom parameter:--cdrom /root/RHEL6.1-20110510.1-Server-x86_64-DVD1.iso
The extra-args parameter is used to pass kernel boot parameters to the OS installer. In this case, since we are connecting to the VM’s serial port, we must use the proper kernel parameters to set it up, just like we would on any server, virtual or not.The extra-args parameter can also be used to specify a kickstart file for non-interactive installations. So if we had a kickstart file we would use:--extra-args="ks=http://my.server.com/pub/ks.cfg console=tty0 console=ttyS0,115200".The OS installation on the VM proceeds as with a physical server, where you provide information such as disk partitions, time zone, root password, etc.Here is another example: Install a RHEL 6.0 x86 VM with 2 VCPUs, 2GB of memory, 15GB disk space, using the default network (private VM network), install from a local iso on the host and use VNC to interact with the VM (must have an X server running):# virt-install \--name vm1 \--ram=2048 \--vcpus=2 \--disk path=/vm-images/vm1.img,size=15 \--cdrom /root/RHEL6.0-20100922.1-Server-i386-DVD1.isoFor more information on all virt-install parameters, refer to the virt-install man page.
If you want several VMs with the same OS and same configuration, I recommend cloning existing VMs rather than installing the OS on each one, which can quickly become a time-consuming & tedious task. In this example, we clone vm1 to create a new VM clone called vm1-clone:1. Suspend the VM to be cloned. This is a requirement since it ensures that all data and network I/O on the VM is stopped.# virsh suspend vm12. Run the virt-clone command:# virt-clone \--connect qemu:///system \--original vm1 \--name vm1-clone \--file /vm-images/vm1-clone.imgThis operation will take 2-3 minutes, depending on the size of the VM.3. When done, you can resume the original VM:# virsh resume vm14. The cloned VM is placed in shutdown mode. To start it:# virsh start vm1-cloneThe cloned VM is an exact copy of the original VM, all VM properties (VCPUs, memory, disk space) and disk contents will be the same. The virt-clone command takes care to generate a new MAC address for the VM clone and updates the proper network controller configuration file (i.e. /etc/sysconfig/network-scripts/ifcfg-em1), thus avoiding duplicate MAC addresses. For more information, refer to the virt-clone man page.
These are some of the commands I use to administer my VMs, and to be honest I don’t use them very often, but they can be handy. As always, for a list of all available commands, your best bet is the virshman page.
List all VMs on a host, running or otherwise:# virsh list --allShow VM information:# virsh dominfo vm1Show VCPU/memory usage for all running VMs:# virt-topShow VM disk partitions:# virt-df vm1Stop a VM (shutdown the OS):# virsh shutdown vm1Start VM:# virsh start vm1Mark VM for autostart (VM will start automatically after host reboots):# virsh autostart vm1Mark VM for manual start (VM will not start automatically after host reboots):# virsh autostart --disable vm1
If you do not have an X server running on your host, connecting to a VMs serial console might be the only way to login to a VM if networking is not available. Setting up access to a VM’s console is no different than in a physical server, where you simply add the proper kernel boot parameters to the VM. For example, for a RHEL VM, append the following parameters to the kernel boot line in /etc/grub.conf and then reboot the VM:console=tty0 console=ttyS0,115200Then, after the VM boots, run in the host:# virsh console vm1
Say you have files in a USB key that you want to copy to your VM. Rather than copying the files to your VM via the network, you can directly attach the USB key (or any USB storage device) to your VM, which will then appear as an additional storage device on your VM. First identify the device name of your USB storage device after you plug it in on the host. In this example, it will be /dev/sdb:# virsh attach-disk vm1 /dev/sdb vdb --driver tap --mode shareable
You can now access the USB storage device directly from your VM at /dev/vdb. When you are done with it, simply detach it from your VM:# virsh detach-disk vm1 vdb
Ok so now I will mention a couple of GUI tools which I admit can be useful when managing and interacting with VMs.
You can easily change VM parameters after creating them, such as memory, VCPUs and disk space.
You can dynamically change the memory in a VM up to what its maximum memory setting is. Note that by default the maximum memory setting in a VM will always equal the amount of memory you specified when you created the VM with the ram parameter in virt-install.So for example if you created a VM with 1 GB of memory, you can dynamically reduce this amount without having to shut down the VM. If you want to increase the memory above 1 GB, you will have to first increase its maximum memory setting which requires shutting down the VM first.In our first example, let’s reduce the amount of memory in vm1 from 1 GB to 512 MB:1. View the VM’s current memory settings:# virsh dominfo vm1 | grep memoryMax memory: 1048576 kBUsed memory: 1048576 kB2. To dynamically set to 512 MB, run:# virsh setmem vm1 524288Value must be specified in KB, so 512 MB x 1024 = 524288 KB3. View memory settings:# virsh dominfo vm1 | grep memoryMax memory: 1048576 kBUsed memory: 524288 kB In our second example, let’s increase the amount of memory in vm1 above from 512 MB to 2 GB:1. In this case we will first need to increase the maximum memory setting. The best way to do it is by editing the VM’s configuration file. Shutdown the VM or you might see unexpected results:# virsh shutdown vm12. Edit the VM’s configuration file:# virsh edit vm1Change the value inside the <memory> tab to:<memory>2097152</memory>Remember that the value is in KB. Save the file and exit.3. Restart the VM from its updated configuration file:# virsh create /etc/libvirt/qemu/vm1.xml4. View memory settings:# virsh dominfo vm1 | grep memoryMax memory: 2097152 kBUsed memory: 524288 kB5. Now you can dynamically change the memory:# virsh setmem vm1 2097152Verify:# virsh dominfo vm1 | grep memoryMax memory: 2097152 kBUsed memory: 2097152 kB
To change the number of virtual CPUs in a VM, change the number in the vcpu tag in the VM’s configuration file. For example, let’s change the number of virtual CPUs to 2:# virsh shutdown vm1# virsh edit vm1<vcpu>2</vcpu># virsh create /etc/libvirt/qemu/vm1.xml
You can always add additional ‘disks’ to your VMs by attaching additional file images. Say that you want to add an additional 10 GB of disk space in your VM, here is what you do:1. Create a 10-GB non-sparse file:# dd if=/dev/zero of=/vm-images/vm1-add.img bs=1M count=102402. Shutdown the VM:# virsh shutdown vm13. Add an extra entry for ‘disk’ in the VM's XML file in /etc/libvirt/qemu. You can look copy & paste the entry for your mail storage device and just change the target and address tags. For example:# virsh edit vm1<disk type='file' device='disk'><driver name='qemu' type='raw' cache='none' io='threads'/><source file='/vm-images/vm1.img'/><target dev='vda' bus='virtio'/><address type='pci' domain='0x0000' bus='0x00' slot='0x04' function='0x0'/></disk><disk type='file' device='disk'><driver name='qemu' type='raw' cache='none' io='threads'/><source file='/vm-images/vm1-add.img'/><target dev='vdb' bus='virtio'/><address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/></disk>
4. Restart the VM from the updated XML configuration file:# virsh create /etc/libvirt/qemu/vm1.xml
When you no longer need a VM, it is best practice to remove it to free up its resources. A VM that is shutdown is not taking up VCPUs or memory, but its image file is still taking up disk space.Deleting a VM is a lot faster than creating one, just a few quick commands. Let’s delete vm1-clone:1. First, shutdown the VM:# virsh shutdown vm1-cloneIf the VM is not responding or fails to shut down, shut it down forcefully:# virsh destroy vm1-clone2. Undefine the VMs configuration:# virsh undefine vm1-clone3. Finally, remove the VM’s image file:# rm /vm-images/vm1-clone.img
Setting up a simple virtualization environment in RHEL 6 is straight-forward and painless. You don’t need a top-of-the line Dell server to get started, any Dell server with sufficient CPU power, memory and disk space is all you need.
In a future article, I will cover more advanced topics such as taking VM snapshots, VM live migration & virtualized-to-virtualized migration (V2V), and get you on your way to becoming a KVM virtualization expert!