User Tools

Site Tools


unix:virtualization:virtlib:quickreference

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
unix:virtualization:virtlib:quickreference [2021/03/07 23:15] rodolicounix:virtualization:virtlib:quickreference [2023/02/26 17:12] (current) rodolico
Line 1: Line 1:
 ====== virt-lib Quick Reference ====== ====== virt-lib Quick Reference ======
 +
 +virtlib stores its configuration for each //domain// (virtual) in /etc/libvirt/qemu*.xml. You should not manually modify anything in this directory, but you can safely do any read-only operation. To edit/export/whatever, the //virsh// command has several options that will help you out.
 +
 +The xml files are, well, xml. The format is documented at https://libvirt.org/formatdomain.html which is guaranteed to either help you, or put you to sleep.
  
 ===== virt-top ===== ===== virt-top =====
Line 11: Line 15:
 Some common commands are: Some common commands are:
   * **list --all** - list all domains virt-lib knows about and show their state (running, off, whatever)   * **list --all** - list all domains virt-lib knows about and show their state (running, off, whatever)
 +  * **vncdisplay //name//** - show vnc display number (not port) assigned
 +  * **dommemstat //name//** - show allocated RAM
 +  * **domstats //name// | grep vcpu.current** - displays number of virtual cpu's assigned
 +  * **domblklist //name//** - show attached block devices
 +  * **domiflist //name//** - list all domain interfaces
   * **start //name//** - starts the domain named **name** (name from list --all)   * **start //name//** - starts the domain named **name** (name from list --all)
   * **reboot //name//**   * **reboot //name//**
Line 20: Line 29:
   * **console domain** - attaches to the serial console of domain assuming the port has been set up.   * **console domain** - attaches to the serial console of domain assuming the port has been set up.
   * Attach/Detach network interfaces (https://computingforgeeks.com/managing-kvm-network-interfaces-in-linux/)   * Attach/Detach network interfaces (https://computingforgeeks.com/managing-kvm-network-interfaces-in-linux/)
-    * attach-interface --domain pxe --type bridge --source br1 --model virtio --config --live +    * attach-interface //domain// bridge //bridge// --config --target vnet3 --mac 00:16:3e:xx:xx:xx --model virtio 
-    * detach-interface --domain pxe --type bridge --mac 52:54:00:47:2f:eb --config+      * This will attach a new bridge to domain //domain//, using //bridge// defined on hypervisor, called vnet3 on the virtual, with the mac address set to whatever you use. 
 +    * detach-interface //domain// --type bridge --mac 00:16:3e:xx:xx:xx --config 
 +  * Remove virtual image (config file only)<code bash> virsh undefine domainname</code> 
 +  * **change-media //name// //drive//** - Insert or Eject a CDROM 
 +    * change-media //name// //drive// --eject 
 +    * change-media //name// //drive// //path/to/image// --insert 
 + 
 +==== Boot from CD ROM ==== 
 + 
 +This is actually not intuitive. I'm going to describe how to do this from a pretty complex setup with no GUI; adjust as needed. In this case, we need to boot the virtual myvirt from a gparted cdrom image located in /media/isos/gparted-live-1.3.0-1-amd64.iso so we can modify the partitions. However, we did not have a cdrom image for the installation, so one has not been created. 
 + 
 +Both assume the virtual has been turned off <code bash>virsh shutdown myvirt</code> 
 + 
 +Some people do not want to manually edit the XML used to configure the virtuals, others prefer it. Some things, like setting a boot menu with a realistic timeout appear to require it; I have not found a way to do it through the virsh command. 
 + 
 +I use the --config flag a lot. If you use this flag, any changes you make are written to the config file and will persist across boots. 
 + 
 +=== without manually editing configuration === 
 +  - First, see if there is a cdrom installed. One simple way is to dump the xml and grep for cdrom<code bash>virsh dumpxml myvirt | grep cdrom</code> 
 +  - Once this is done and you have a cdrom, you should be able to tell which one it is with<code bash>virsh domblklist myvirt</code> which shows the block devices attached to myvirt. 
 +  - Mount your image on the CDROM 
 +    - If you have one and you know which drive it is<code bash>virsh change-media dd-113-virt hdb /media/xen-store/isos/gparted-live-1.3.0-1-amd64.iso --insert --config</code>This is assuming the drive was hdb 
 +    - If you do not have a CDROM, use<code bash>virsh attach-disk myvirt /media/xen-store/isos/gparted-live-1.3.0-1-amd64.iso hdb --driver file --type cdrom --mode readonly --config</code>This assumes hdb did not show up as used in your domblklist above. 
 + 
 +=== I'm not afraid to edit the config === 
 +  - Edit the config file with <code bash>virsh edit myvirt</code> 
 +  - Look for something like //<disk type='file' device='cdrom'>// 
 +  - If it exists, add <code xml><source file='/media/isos/gparted-live-1.3.0-1-amd64.iso'/></code> 
 +  - Look for the section //<os>//, generally near the top and add lines if necessary 
 +    - <code xml><bootmenu enable='yes' timeout='5000'/></code> 
 +    - This will enable the boot menu, and have a wait of 5 seconds (5000 milliseconds) for you to choose. 
 +  - If you want, change the boot order by adding <code xml><boot dev='cdrom'/></code> //above// the boot entry if it is not already set. 
 + 
 +=== Starting the virtual === 
 +  - Start the virtual with <code bash>virsh start myvirt</code> 
 +  - Immediately make the VNC connection (if you followed the manual edit, you have 5 seconds) 
 +  - When prompted, press ESC to choose the CDROM drive (assuming you didn't set it as the default) 
 +=== Cleanup === 
 +Do the following if you want to remove the CDROM when you're done. Note: this is likely required if you set the CDROM as the primary boot device. 
 +<code bash>virsh change-media myvirt hdb --eject --config</code> 
 + 
  
 ==== Remove a network from the entire system ==== ==== Remove a network from the entire system ====
Line 41: Line 91:
 //virt-install// does not allow you to use two ISO's as cdrom images by default. This can be a problem when you want to do a new install, but want to use virtio on Windows (which requires a second ISO. You may be able to do a change-media, but the following, taken from [[https://superuser.com/questions/147419/using-virt-install-to-mount-multiple-cdrom-drives-images //virt-install// does not allow you to use two ISO's as cdrom images by default. This can be a problem when you want to do a new install, but want to use virtio on Windows (which requires a second ISO. You may be able to do a change-media, but the following, taken from [[https://superuser.com/questions/147419/using-virt-install-to-mount-multiple-cdrom-drives-images
 ]] will do the trick with only one extra step. ]] will do the trick with only one extra step.
 +
 +**Note**: You may have to manually start the virtual several times during the installation. Windows reboots 2-3 times during an installation, and sometimes this is not handled well by virsh.
  
   - Add the --print-xml flag to virt-install, and pipe the output to a file.   - Add the --print-xml flag to virt-install, and pipe the output to a file.
Line 46: Line 98:
     - Change the drive letter     - Change the drive letter
     - Change the target to point to virtio-win.iso     - Change the target to point to virtio-win.iso
-  - run virsh create name_of_xml_file to begin the installation+  - run <code bash>virsh create name_of_xml_file</code> to begin the installation 
 +  - After installation is complete, install the remainder of the virtio drivers for maximum efficiency 
 +  - To permanently define after installation 
 +    - Edit XML and remove the second CDROM (if desired) 
 +    - <code bash>virsh define name_of_xml_file</code>
  
  
Line 91: Line 147:
 I did this with a Windows 10 installation on my workstation. Prior to using virtio, it would take, literally, 5-7 minutes after boot before I could do anything, and it was very sluggish after that. Once I used virtio, it was almost bare hardware speeds. I did this with a Windows 10 installation on my workstation. Prior to using virtio, it would take, literally, 5-7 minutes after boot before I could do anything, and it was very sluggish after that. Once I used virtio, it was almost bare hardware speeds.
  
-==== Links ====+===== Shutdown and restart of Windows guests ===== 
 + 
 +==== Problems shutting down with //virsh// ==== 
 +After installing the win-virtio package, you should see QEMU Guest Agent running as a service. With this running, you can use qemu-guest-agent to manage shutdown and reboot. This is much more reliable than using ACPI. 
 + 
 +**Important** Shut down the virtual before doing the following. 
 + 
 +Edit the guest 
 +<code bash> 
 +virsh edit DOMAIN 
 +</code> 
 + 
 +Place the following block under the <devices> section. I usually put it after the <console> sections, but just so long as it is in the <devices> section of the config. Virsh will rearrainge it for you anyway. 
 + 
 +<code xml> 
 +<channel type="unix"> 
 +  <source mode="bind"/> 
 +  <target type="virtio" name="org.qemu.guest_agent.0"/> 
 +</channel> 
 +</code> 
 + 
 +Start the virtual back up. Once that is done you can use the following commands much more reliably. 
 +<code bash> 
 +virsh shutdown DOMAIN 
 +virsh reboot DOMAIN 
 +</code> 
 + 
 +This also allows you to execute //qemu-agent-command// from within //virsh//, but this is strongly discouraged unless you have researched what is going on. Any changes you make can cause instability in libvirt, since they are bypassing virsh. Kind of like using hdparm, where you can do Really Bad Things to your hard disk if you don't know for sure what you are doing. Do what you want, but be careful. 
 + 
 +==== Using save/restore ==== 
 + 
 +An alternative is to use //virsh save// to shut down and //virsh restore// to recover. In this case, use 
 +<code bash> 
 +virsh save domainname /path/to/save/image 
 +</code> 
 +to save the image. It will suspend the virtual, then save processor/memory/whatever to the file /path/to/save/image. 
 + 
 +Once this is done, you can do whatever you needed to do, then use the following command to restore it. 
 +<code bash> 
 +virsh restore /path/to/save/image 
 +</code> 
 + 
 +==== Windows servers will not restart ==== 
 + 
 +I'm having a problem with Windows virtuals not rebooting. When you issue the restart command, they shut off and don't come back up. As a band aid, I have a script running on the hypervisor with a cron job, every 5 minutes. 
 + 
 +This script has been tested on our machines, but I'm sure there are some issues with it. Just do a  
 +<code bash> 
 +virsh list --all 
 +</code> 
 +and select the domains you want to ensure are running all the time. Place them in the array that has DOMAIN1 and DOMAIN2 (ie, replace DOMAIN1 with your first choice, etc...). 
 + 
 +When called, checkVirtuals will look for each of the domains and see if they are running (using //virsh list//). If they are not running, it will place a flag file in /tmp/DOMAIN.down. The next time it is run, it will note the domain is still not down, and the flag file exists, so it will start the domain (using //virsh start DOMAIN// and delete the flag file. 
 + 
 +I call this every 5 minutes from cron, thus, the max downtime will be 10 minutes, with an average of 5. 
 + 
 +**WARNING:** Remember this is running. If you need to take a virtual down for some reason, as long as this script is running, it will blindly go ahead and restart it. 
 + 
 +<code perl checkVirtuals> 
 +#! /usr/bin/env perl 
 + 
 +use strict; 
 +use warnings; 
 + 
 +my @servers = (  
 +   'DOMAIN1', 
 +   'DOMAIN2' 
 +   ); 
 +    
 +my $virsh = '/usr/bin/virsh start '; 
 + 
 +my $output = `virsh list`; 
 + 
 +foreach my $server ( @servers ) { 
 +   if ( $output =~ m/$server/ ) { 
 +      unlink "/tmp/$server.down" if  -e "/tmp/$server.down"; 
 +   } else { 
 +      if ( -e "/tmp/$server.down" ) { 
 +         print "$server has been down for a while, starting back up\n"; 
 +         `$virsh $server`;  
 +         unlink "/tmp/$server.down"; 
 +      } else { 
 +         `touch /tmp/$server.down`; 
 +      } 
 +   } 
 +
 + 
 +1; 
 + 
 +</code> 
 + 
 +===== Replacing Network Interfaces ===== 
 + 
 +Sometimes you need to undefine and redefine a network interface. This is actually fairly simple to do. 
 + 
 +<code bash> 
 +# get a list of all network interfaces in domain 
 +virsh domiflist domain 
 +# remove the one you want. Use values (type, mac) from above command 
 +virsh detach-interface domain --type bridge --mac ##:##:##:##:##:## --config 
 +# redefine it. Use values from above command, or change as needed 
 +virsh attach-interface domain --type bridge --model virtio --source br_private --mac ##:##:##:##:##:## -config 
 +</code> 
 + 
 +In this example, //domain// is the name of the domain to be worked on. I use domiflist to get information about the network interfaces as I need the type and mac. The --config makes it permanent, writing it to the config file. 
 + 
 +In my case, I had built a Windows domain without using virtio, and I wanted to change it. This was actually the simplest way I found to do it. 
 + 
 +===== Naming Network Links ===== 
 + 
 +to give a domain interface a static name when it is run, you can add the following to the interface definition. 
 +<code xml><target dev='dom113'/></code> 
 +NOTE: the name must not begin with 'vnet', 'vif', 'macvtap', or 'macvlan', which are prefixes reserved by libvirt and certain hypervisors. See 
 +https://libvirt.org/formatdomain.html#overriding-the-target-element 
 +<code xml> 
 +   <interface type='bridge'> 
 +      <mac address='00:16:3e:6b:f4:da'/> 
 +      <source bridge='br_lan'/> 
 +      <target dev='dom113'/> 
 +      <model type='virtio'/> 
 +      <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/> 
 +    </interface> 
 +</code> 
 + 
 +You can now do things like <code bash>ifconfig dom113</code> or <code bash>ip a show dom113</code> 
 + 
 +===== Adding USB Drive Image ===== 
 + 
 +Building and using a USB Thumbdrive image is a little weird, but it can be done. See [[software:multibootusb|]] for an example of one way I did it. 
 + 
 +===== Links =====
  
   * https://www.caretech.io/2017/10/18/converting-windows-vm-hard-disk-interface-to-virtio-with-proxmox-ve-5/   * https://www.caretech.io/2017/10/18/converting-windows-vm-hard-disk-interface-to-virtio-with-proxmox-ve-5/
   * https://docs.fedoraproject.org/en-US/quick-docs/creating-windows-virtual-machines-using-virtio-drivers/   * https://docs.fedoraproject.org/en-US/quick-docs/creating-windows-virtual-machines-using-virtio-drivers/
   * https://linuxhint.com/install_virtio_drivers_kvm_qemu_windows_vm/   * https://linuxhint.com/install_virtio_drivers_kvm_qemu_windows_vm/
 +  * https://serverfault.com/questions/672253/how-to-configure-and-use-qemu-guest-agent-in-ubuntu-12-04-my-main-aim-is-to-get#691616
 +  * https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/6/html/virtualization_administration_guide/sect-qemu_guest_agent-running_the_qemu_guest_agent_on_a_windows_guest
 +  * https://libvirt.org/formatdomain.html
 +
unix/virtualization/virtlib/quickreference.1615180523.txt.gz · Last modified: 2021/03/07 23:15 by rodolico