- Users - libvirt List Archives

[libvirt-users] Forward routed network bridge on system's vlan
by RaSca 30 May '11

30 May '11

Hi all, I created a two node cluster that manages virtual machines with two servers connected via a cross cable on the network 10.0.0.0/24. I want that machines that runs on different servers in the network 172.16.0.0/24 can see all the others. To make this possible I've configured a vlan on each server: ... ... 5: eth1.111@eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP link/ether 5c:d9:98:b1:4c:4d brd ff:ff:ff:ff:ff:ff inet 172.16.0.0/24 brd 172.16.0.255 scope global eth1.111 inet6 fe80::5ed9:98ff:feb1:4c4d/64 scope link valid_lft forever preferred_lft forever ... ... And created a "lan" network, via virsh: <network> <name>lan</name> <uuid>8ac034c8-553d-b995-5a80-c1c711cb088c</uuid> <forward dev='eth1.111' mode='route'/> <bridge name='virbr1' stp='on' delay='0' /> <ip address='172.16.0.2' netmask='255.255.255.0'> </ip> </network> but with this setup "lan" never comes up because, for some reason, the network is locked: # virsh net-start lan error: Failed to start network lan error: internal error Network 172.16.0.1/255.255.255.0 is already in use by interface eth1.111 What is not clear to me is why, if I turn down the eth1.111 device, start the lan network, turn eth1.111 up and then force the brctl addif command, everything works: # ifdown eth1.111 Removed VLAN -:eth1.111:- # virsh net-start lan Network lan started # ifup eth1.111 Set name-type for VLAN subsystem. Should be visible in /proc/net/vlan/config Added VLAN with VID == 111 to IF -:eth1:- # brctl addif virbr1 eth1.111 # brctl show bridge name bridge id STP enabled interfaces br0 8000.6c626dd7518e no eth0 virbr0 8000.000000000000 yes virbr1 8000.5cd998b14c4d yes eth1.111 Why manually is possible to make things work? What am I ignoring? Thanks a lot, -- RaSca Mia Mamma Usa Linux: Niente è impossibile da capire, se lo spieghi bene! rasca(a)miamammausalinux.org http://www.miamammausalinux.org

1 0

[libvirt-users] libvirt boots from Knoppix-CD but not Debian-CD
by Christian Brandt 30 May '11

30 May '11

Debian Squeeze, libvirt 0.8.3, qemu-kvm 0.15.5 After configuring the cd image the bios screen shows: (tested with virsh, virt-install and virt-manager) Starting SeaBIOS (version 0.5.1-20110523_174945-brahms) Booting from CD-Rom 65MB medium found (bzw. whatever size) Boot failed: Could not read from CDROM (code 000c) ... the cd images (debian6-cd1, debian6-netinst, debian6-buisnesscard) have the correct md5sum an can be mounted, they simply do not boot. On the other hand, a recent knoppix ISO works well. Tested countless times. All access rights are 777. Only special feature: the host is a minimal installation done with debootstrap and contains not much else than libvirt, virt-manager and lxterminal. I am out of wit. Any ideas? -- Christian Brandt life is short and in most cases it ends with death but my tombstone will carry the hiscore

1 2

[libvirt-users] kvm.h not found
by --[ UxBoD ]-- 28 May '11

28 May '11

Hello all, Am attempting to compile the latest version of libvirt against a custom kernel 2.6.32.40 under CentOS 5.6 and see this in the configure output: checking linux/kvm.h usability... no checking linux/kvm.h presence... no checking for linux/kvm.h... no I have tried to change where it is looking using: CPPFLAGS="-I /home/compuser/linux-2.6.32.40/include" ./configure but end up with the same result. What am I doing wrong ? -- Thanks, Phil

2 1

[libvirt-users] polkit error when starting virt-manager on fedora14
by Alex 27 May '11

27 May '11

Hi, I have the following error in syslog when starting virt-manager and trying to connect: May 25 16:39:33 sage libvirtd: 16:39:33.525: error : remoteDispatchAuthPolkit:3846 : Policy kit denied action org.libvirt.unix.manage from pid 27509, uid 500, result: 512 I can manually start the kvm guest using virsh, but why can't virt-manager start them? This is on fedora14 x86_64: # rpm -qva|egrep 'virt|polkit|qemu' polkit-qt-0.99.0-2.fc14.x86_64 qemu-img-0.13.0-1.fc14.x86_64 python-virtkey-0.50-8.fc14.x86_64 qemu-kvm-0.13.0-1.fc14.x86_64 polkit-0.98-5.fc14.x86_64 polkit-gnome-0.97-4.fc14.x86_64 libvirt-python-0.8.3-9.fc14.x86_64 virt-viewer-0.2.1-1.fc13.x86_64 gpxe-roms-qemu-1.0.1-3.fc14.noarch virt-top-1.0.4-3.fc13.x86_64 libvirt-client-0.8.3-9.fc14.x86_64 virtuoso-opensource-6.1.2-1.fc14.x86_64 qemu-common-0.13.0-1.fc14.x86_64 polkit-desktop-policy-0.98-5.fc14.noarch redland-virtuoso-1.0.12-1.fc14.x86_64 python-virtinst-0.500.6-1.fc14.noarch libvirt-0.8.3-9.fc14.x86_64 qemu-system-x86-0.13.0-1.fc14.x86_64 virt-manager-0.8.7-2.fc14.noarch Please let me know what other information I can provide to fix this problem. I haven't rebooted yet, but I suspect at least that would fix it? Thanks, Alex

1 1

[libvirt-users] Failed to connect to Xen 4.1 using virsh
by YAO 27 May '11

27 May '11

Hi, I installed xen 4.1 and libvirt 0.9.1 on ubuntu 10.04 x86 when I tried "*virsh -c xen:/// list*", the terminal give me error: *error: unable to connect to '/var/run/libvirt/libvirt-sock', libvirtd may need to be started: No such file or directory* *error: failed to connect to the hypervisor* But when I tried "*sudo virsh -c xen:/// list*", there's different error: *error: unable to connect to 'localhost:8000': Connection refused* *error: failed to connect to the hypervisor* That really confused me. Do I need to start libvirtd like "*sudo libvirtd*"? cause "*virsh -c test:///default list*" is OK without it. I have started xencommons and xend, so I really don't know what I missed... Thanks, Yao

1 0

[libvirt-users] Replace library WinXPs by virtual WinXPs as on a linux Fedora system?
by Rob 25 May '11

25 May '11

Hi, The current situation: * About 500 university library PCs run WinXP as the main OS. I would like to achieve the following: * Replace the WinXP by a linux system (Fedora?) with virtualizationsoftware (KVM?) and have WinXP as a guest OS, in such a way that the users immediately boot into WinXP without noticing the virtualization. * Over the network I should still be able to access the host Fedora OS. Is this a feasible idea? If yes, can you give me some clues how to start? Thank you! Rob.

1 0

[libvirt-users] Replace library WinXPs by virtual WinXPs as on a linux Fedora system?
by Rob 25 May '11

25 May '11

Hi, The current situation: * About 500 university library PCs run WinXP as the main OS. I would like to achieve the following: * Replace the WinXP by a linux system (Fedora?) with virtualizationsoftware (KVM?) and have WinXP as a guest OS, in such a way that the users immediately boot into WinXP without noticing the virtualization. * Over the network I should still be able to access the host Fedora OS. Is this a feasible idea? If yes, can you give me some clues how to start? Thank you! Rob.

1 0

[libvirt-users] Xen and live migration with VIR_MIGRATE_PERSIST_DEST
by Thibault VINCENT 25 May '11

25 May '11

Hi, I'm having a hard time to figure how to perform Xen live migration using the python API with the same rate of success than virt-manager. # virsh version Compiled against library: libvir 0.8.3 Using library: libvir 0.8.3 Using API: Xen 3.0.1 Running hypervisor: Xen 3.2.0 The goal is to have the migration done in only one call, including a definition in the remote xend and undefinition from the source, which requires the VIR_MIGRATE_PERSIST_DEST and VIR_MIGRATE_UNDEFINE_SOURCE flags. Here's the code doing the migration: ---------- flags = (libvirt.VIR_MIGRATE_LIVE ^ libvirt.VIR_MIGRATE_PERSIST_DEST ^ libvirt.VIR_MIGRATE_UNDEFINE_SOURCE) uri = 'xenmigr://remote.host.com:8002' domain_obj.migrate(con_handle, flags, None, uri, 0) ---------- There are various problems that occur randomly, such as the VM completely disappearing 5 seconds after a succesful migration, but the main issue is with VIR_MIGRATE_PERSIST_DEST that doesn't seem to have any effect. The domU is not defined in xend and will vanish when destroyed or on hypervisor reboot. Could that be a known problem or related to a bug ? I'm really out of ideas. Any help appreciated, Thank you ! -- Thibault VINCENT - IT SmartJog S.A.S. - Groupe TDF - Pôle multimédia 27 Bd Hippolyte Marques, 94200 Ivry-sur-Seine, France Phone: +33.1.58.68.62.38 / Fax: +33.1.58.68.60.97

1 0

[libvirt-users] Java Bindings for libvirt
by Atul Gosain 25 May '11

25 May '11

Hi I looked at the java bindings for libvirt at http://libvirt.org/sources/java/ . Latest version here is 0.4.7 whereas the latest version of libvirtd is 0.9.1 . Does the java binding use latest libvirtd ? If not, can it be created easily ? Also, is there any plan to support HyperV support in future ? Thanks Atul

3 3

[libvirt-users] Live Migration UDP implementation
by prakash hr 24 May '11

24 May '11

Hi all, For my academic project to analyse the Performance of transport protocols in Live Migration of Virtual Machines.I have configured the live migration using tcp and I succeed the live migration. Now for my further remaining project work I need to do the same thing for the UDP implemataiton.When I changed the command in virsh, (My fianl command in TCP is virsh migrate --live ubuntu21 qemu+ssh://172.16.17.23/systemtcp:// 172.16.17.23:444) to udp it is not workng. It is given that qemu embeds user mode network stack which includes tcp,dhcp,udp.etc within the wemu process. Its giving the error no API found for UDP. ---Now i want to implement the patch file for migration using UDP if possible. ---Where I should change the API in order to perform the live migration using UDP. ---Is any patch file available so that modifying that patch file will perform the Live Migration using UDP. ---If implementation using not possible with UDP please give me the details why that can not be implemented and is any proofs for that reasons are appreciated. --Is there any other transport layer protocols can be used instead of TCP during Live migration.? --How about using RUDP instead of TCP. Thanking You in advance. Thanks and Regards Prakash H.R M.Tech Student at RVCE, Bangalore +91-8904301073

2 1

[libvirt-users] getting live migration to work..
by Jelle de Jong 24 May '11

24 May '11

Hello everybody, I am trying to get migration to work with the pacemaker VirtualDomain RA. However I hit some problems and went back to the basic. I also can't get virsh to do the migration so I am kind of stuck and can use some advice. # connection and general info and the migration tests http://paste.debian.net/117736/ # capabilities http://paste.debian.net/117745/ Thanks in advance, Kind regards, Jelle de Jong

1 1

[libvirt-users] Improving video for Windows7 kvm guest on fc14
by Alex 23 May '11

23 May '11

Hi, I have an existing windows7 x86_64 kvm guest running on fedora14 x86_64 with a Radeon HD5700 at 1680x1050, and would like to improve the resolution of the Windows guest. How can I change the Windows driver to something more capable than "standard vga"? It currently reports there is only 16MB of RAM at available at 1440x900. How can I increase this? I've tried using virsh or editing the xml directly, but it doesn't seem to be reflected in the guest. Some of the configuration options I've read in the documentation don't work with fedora14 and virt-manager, such as the acceleration options. This is my current configuration: <video> <model type='vmvga' vram='192000' heads='1'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x02' function='0x0'/> </video> How can I more effectively use the video card's capabilities in the Windows guest? Thanks, Alex

1 0

[libvirt-users] Migrate doesn't support copy-storage-all
by Abhishek Gupta 21 May '11

21 May '11

While using --copy-storage-all option while doing live migration I am getting following error : error: command 'migrate' doesn't support option --copy-storate-all I am using kvm hypervisor. Does that mean kvm doesn't support the option or libvirt doesn't provide support for the same with kvm as the hypervisor? -- Abhishek Student, IIT Delhi

1 0

[libvirt-users] Blank console and no boot
by Francesc Guasch 19 May '11

19 May '11

Hi. I was cloning a domain and now I can't boot the original nor the clone domain. Currently when I start one of those domains I can see only a blank screen. I am not sure what went wrong cloning. I was trying to use kpart to mount the new image and change info in the filesystem like hostname, ip and so. Anyway when I start the new or old domain: # virsh start --console thedomain Domain thedomain started Connected to domain thedomain Escape character is ^] And that's all.I see nothing with libvirt graphical console, nothing in the logs but this, that I guess it's the usual startup info: LC_ALL=C PATH=/usr/local/sbin:/usr/local/bin:/usr/bin:/usr/sbin:/sbin:/bin QEMU_AUDIO_DRV=none /usr/bin/kvm -S -M pc-0.12 -enable-kvm -m 512 -smp 1,sockets=1,cores=1,threads=1 -name .... ... This is qemu-kvm 0.12.5+noroms-0ubuntu7.1 on ubuntu server 10. Any hints so I can check what could have go wrong ? I am concerned about not being able to see what is going on. I guess there must be a way to start the domains and watch the boot process somehow. Is this possible ? thank you for you time

1 0

[libvirt-users] Trying to re-compile libvirt package on Ubuntu Lucid
by Lars Nordin 18 May '11

18 May '11

I'm trying to rebuild Ubuntu's libvirt packages to turn on ESX support. I've downloaded the Ubuntu source package for Lucid (updated Lucid package) dget -u http://archive.ubuntu.com/ubuntu/pool/main/libv/libvirt/libvirt_0.7.5-5ubun… I try and re-compile dpkg-buildpackage -rfakeroot -b DEB_BUILD_OPTIONS="--with-esx" fakeroot debian/rules binary and I get this error: ... testing with corrupted config: log_level Skipping test (due to Ubuntu buildd FTBFS): running libvirtd with a valid config file (2 seconds) FAIL: daemon-conf PASS: define-dev-segfault ... ==================== 1 of 37 tests failed ==================== make[4]: *** [check-TESTS] Error 1 make[4]: Leaving directory `/exp/h1/staff/lnordin/libvirt-pkg/libvirt-0.7.5/tests' make[3]: *** [check-am] Error 2 make[3]: Leaving directory `/exp/h1/staff/lnordin/libvirt-pkg/libvirt-0.7.5/tests' make[2]: *** [check-recursive] Error 1 make[2]: Leaving directory `/exp/h1/staff/lnordin/libvirt-pkg/libvirt-0.7.5/tests' make[1]: *** [check-recursive] Error 1 make[1]: Leaving directory `/exp/h1/staff/lnordin/libvirt-pkg/libvirt-0.7.5' make: *** [debian/stamp-makefile-check] Error 2 Anyone gone through this exercise before and would know how to get the packages re-built?

1 0

[libvirt-users] Howto set the option script=no for a tap device in the xml file
by Christian Spann 18 May '11

18 May '11

Hello, I am trying to run a virtual machine with qemu-kvm and libvirt as a user. In order to start the virtual machine I setup a tap device manually, add it to the bridge and activate it via ifconfig. But I cannot tell libvirt to not try to add the definded tap device to the bridge by itself. It always fails telling me it cannot add tap0 to br0. Thats the xml snippet for defining the network: <interface type='bridge'> <mac address='54:52:00:38:75:dd'/> <source bridge='br0'/> <target dev='tap0'/> <model type='virtio'/> </interface> Here I would like do define "script=no" to get an output like: -net tap,vlan=0,ifname=tap0,script=no when libvirt launches the vm. Regards, Christian -- Christian Spann - http://www-vs.informatik.uni-ulm.de/~spann

2 2

[libvirt-users] (no subject)
by Abhishek Gupta 17 May '11

17 May '11

Hi, I am using libvirt and virsh for my application. I am trying to manage snapshots for my infrastructure. All virsh commands are working fine like snapshot-create, delete, list, revert, current. But when I try to do the same using libvirt, I get out of memory error. Here are the commands that i execute on my host in a python shell... import libvirt conn = libvirt.open("qemu:///system") dom = conn.lookupByName("vmtest") *snap = dom.snapshotCreateXML(None,0);* The error is as follows libvir: Remote error : out of memory Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/lib/python2.6/dist-packages/libvirt.py", line 669, in snapshotCreateXML if ret is None:raise libvirtError('virDomainSnapshotCreateXML() failed', dom=self) libvirt.libvirtError: out of memory

2 2

[libvirt-users] Looking for a contact
by Dale Amon 17 May '11

17 May '11

I'm curious if there is anyone here from linode.

1 0

[libvirt-users] libvirt connection timeout
by Antoine COETSIER 16 May '11

16 May '11

Hi all, I would like to set a short connection timeout (only some seconds) when using libvirt in my Python program instead of the long default one. I found the C function: virEventAddTimeoutFunc() in the C libvirt API here: http://libvirt.org/html/libvirt-libvirt.html#virEventAddTimeoutFunc and eventInvokeTimeoutCallback(timer, callback, opaque) in libvirt.py around the line#150 but I do not know how to use it. I did not find any example on the net. I tried this but I get a segmentation fault: :-( import libvirt def timeout_cb_d(): print 'Timeout !' try: # try to set the libvirt timeout to 2 seconds: t = libvirt.eventInvokeTimeoutCallback(2, timeout_cb_d, "from dom0_class") except: ... Does anyone can give me a working example please? Antoine Coetsier

3 3

[libvirt-users] (no subject)
by Abhishek Gupta 16 May '11

16 May '11

Hello, In the documentation section of Python API Website it is written that the generation of python API and documentation is automated using generator.py file. This should produce the necessary files alongwith the documentation in libvirtclass.txt which should be present in docs or python folder. I executed generator.py file but was not able to find the libvirtclass.txt file to get the documentation. The easiest way I could think of was to manually browse the libvirt.py file for the available functions. Is there any other way of doing the same? -- Abhishek

2 1

[libvirt-users] (no subject)
by Abhishek Gupta 15 May '11

15 May '11

Hello, I am using python bindings of libvirt. When I use the function listDomainsID( ) it successfully returns me the list of the ids of my current virtual machines, but when I call function listDefinedDomains( ) it returns me an empty list. Is this some kind of bug because as per my understanding listDefinedDomains should return a superset of vm's as returned by listDefinedDomains(). I am using 0.8.8 version of the libvirt. -- Thanks Abhishek

2 2

Re: [libvirt-users] Problem after restarting libvirt
by Nikita A Menkovich 15 May '11

15 May '11

I could not find that cgroups are not persistent now. And there is no such information in changelog. Could you show me commit where it was changed? On 9 May 2011 15:50, Nikita A Menkovich <menkovich(a)gmail.com> wrote: > It is really bad. Ok, I will write a script, that will set up cgroups > for every domain. > > On 9 May 2011 10:44, Osier Yang <jyang(a)redhat.com> wrote: >> 于 2011年05月05日 21:32, Nikita A Menkovich 写道: >>> >>> After restarting libvirt cgroups for VMs are not recovered. >>> Is there available to restore cgroup state after restart? >>> >> except "cpu.shares" (Since 0.9.0), all other cgroup related stuffs >> have no persistent XML yet, so they can't be recovered. >> >> Regards >> Osier >> > > > > -- > Nikita A Menkovich > http://libc6.org/ > JID: menkovich(a)gmail.com > Tel: +7 (921) 423-96-48 > -- Nikita A Menkovich http://libc6.org/ JID: menkovich(a)gmail.com Tel: +7 (921) 423-96-48

2 1

[libvirt-users] Booting from scsi
by bhasker＠unixindia.com 14 May '11

14 May '11

Hi all, I am starting up to use libvirt with virsh and virt-install. I used virt-install successfully and have a running linux system with raw disk exported to the virtual machine as scsi. (the whole of qemu + libvirt + virtmanager is compiled and installed on /usr/local from source) After the virtual OS installation (which succeeds) I am not able to boot into the OS from the harddisk. Following are some excerpts from my xml file. Please get back to me if you need any more information <os> <type arch='x86_64' machine='pc-0.14'>hvm</type> <boot dev='hd'/> <bootmenu enable='yes'/> </os> I also tried removing the boot parametes from <os> and added <boot order='1'/> in the scsi disk section. Still I am not able to use the scsi disk to boot (my qemu-system-x86_64 however has the boot=on option) <disk type='block' device='disk'> <driver name='qemu' type='raw'/> <source dev='/dev/venkatb1/deb'/> <target dev='sda' bus='scsi'/> <address type='drive' controller='0' bus='0' unit='0'/> </disk> I could not find any more information or any posts where someone else had faced this issue. I presume I am doing something wrong here. Can someone help please ? thanks Bhasker C V

1 1

[libvirt-users] Problems using virt-install to create KVM CentOS 5.6 guest
by Emmanuel Noobadmin 13 May '11

13 May '11

I was having problems creating a KVM CentOS 5.6 Initially the host was CentOS 5.6 but I switched to Scientific Linux 6 during the process of trying to solve it. I was able to install a working guest in the end by manually writing an XML and using virsh. However, as it shouldn't be so complicated, Gleb over at the KVM list asked me to share this with the libvirtd ML. I don't know it's an old bug, since current libvirt seems to be several sub-versions ahead of mine so decided to post to the user list first in case it's a known bug. Hopefully somebody will be able to pinpoint what went wrong. Host system: SL 6.0 with libvirtd (libvirt) 0.8.1 Network: 2 physical NICs, eth0 bridged to br0 for WAN connectivity, eth1 is for internal use Installation medium: CentOS 5.6 DVD ISO Original problem virt-install would simply not work, I get a blank console with nothing happening and qemu-kvm will be 100% load on one core. This was eventually traced to virt-install not being able to mount/use the ISO. Mounting the ISO as loopback allowed the installation to proceed. However, the guest was not able to see the installation "cdrom". With httpd running off the host, and tested accessibly via Internet, a network install did not work either. After much frustration, I followed the documentation at libvirt website to write my own XML file. The result of this was the discovery that 1. permissions on initrd.img and vmlinux was not allowing virsh access - copying these files to disk and adjusting the xml allowed installation to proceed. virt-install appears to do this on the fly. 2. virt-install guest could not see any disk. The anaconda installer could not find any CDROM or hard disk. - Part of the problem seems to be that virt-install did not save the source for the installation medium, the generated XML did not have the source information under the <disk> element. I specified this in my manual XML and the CentOS guest was able to see the "cdrom" and install properly. As I did not have any problem with missing hard disks with my manually written XML so I'm not sure what went wrong with virt-install's 3. Can't install via network - This was traced to networking not working at all. Even in my manual XML, networking did not work although the guest could install from the disk. Eventually, this was solved by adding the <model type='virtio' /> to the device definition.

1 1

[libvirt-users] point-to-point connection between vm and host
by Jorge Somavilla 12 May '11

12 May '11

Hi! hopefully someone can give me a hint with this: I need to create a vm with a point-to-point connection to the host using a /30 subnet. Right now i have this in the vm's xml for libvirt: <interface type="network" name="eth0" onboot="yes"> <source network="default"/> <mac address="02:fd:00:00:01:00"/> </interface> Which brings up an interface in the vm that is configured later by a a daemon, getting to this: vnx@r2:~$ ifconfig eth0 eth0 Link encap:Ethernet HWaddr 02:fd:00:00:01:00 inet addr:10.250.0.2 Bcast:10.250.0.3 Mask:255.255.255.252 While in the host the other interface is created: root@tau:/home/jorge/workspace/vnx# ifconfig r2-e0 r2-e0 Link encap:Ethernet HWaddr 0a:81:b6:3c:24:15 inet addr:10.250.0.1 Bcast:10.250.0.3 Mask:255.255.255.252 However there is no connectivity between them, probably because of some mistake in the <interface> tag, perhaps a wrong 'type' attribute, but i haven't found out yet. I wouldn't want to create a bridge and connect both interfaces (vm's and host's) to the bridge, but to have this ptp, if possible. How would i go about doing that? thanks! Jorge

2 3