June 2012 - Users - libvirt List Archives

[libvirt-users] Can I update the DNS configuration of an existing network?
by Kees van Reeuwijk 26 Jun '12

26 Jun '12

Hello, As part of a research project on cloud management, I am developing software to create on-demand networks between VMs on different hosts. The deployment of the VMs themselves is done by OpenNebula, but for the management of the networks I have to bypass OpenNebula, and I talk directly with libvirt. One of the features of our network management is that we allow network connections to be created and destroyed at any moment in the life of a VM. This means that we use hotplugging. It also means that we want a VM to discover its network parameters for a particular NIC using DHCP and DNS. I have something that mostly works now: I create an XML for each network, I can hot-plug new NICs in the VM using libvirt, and DHCP on the new interface works. However, DNS does not work properly. I have to create a mapping between the MAC, IP address, and the name of a VM at the moment I create the network XML. Setting the MAC <-> IP mapping at that time is fine, but fixing the mapping between IP address and name at that time is not reasonable. I do not know what name will be used before I add a new machine, and during the lifetime of the network there may even be different names associated with a particular IP address. Is there a way I can change/update the DNS configuration of a network after it has been created? I could simply create a new network XML file and stop/start the network, but I suspect that is too disruptive. Is there a less intrusive way to do this? Another thing I would very much like to do is to talk to libvirt directly from my management software (written in Java), rather than through the command line. I know there is a Java API for libvirt, but I have not been able to find any up-to-date documentation on it. Is the functionality that I need accessible through the Java API? Is there enough documentation of the API for this, and where is it? -- Dr. Ir. Kees van Reeuwijk, Vrije Universiteit Amsterdam

1 0

[libvirt-users] USB Host Controllers
by Michael Hierweck 26 Jun '12

26 Jun '12

Hi, it is possible to define a domain along with several usb host controllers, e.g. <controller type='usb' index='0' model='ich9-ehci' /> <controller type='usb' index='1' model='ich9-uhci' /> How to distinguish between those controllers/busses when adding input or host devices such as <input type='tablet' bus='usb'/> <hostdev mode='subsystem' type='usb' managed='no'> <source> <vendor id='...'/> <product id='...'/> </source> </hostdev> Without explicit assignment either every device seems to be attached to the uhci controller or the domain refuses to start because the tablet is attached to the ehci controller implicitly, depending on the order of the controller definition (index). Thanks in advance Michael -- EDV-Serviceteam Annika & Michael Hierweck GbR Egerstraße 53, 44225 Dortmund (Germany) http://www.edv-serviceteam.net

2 1

[libvirt-users] Is it possible to in live migration?
by GaoYi 25 Jun '12

25 Jun '12

Hi all, In live migration using shared memory, there would be a following step to input the password after typing: migrate --live vm_name qemu+ssh://Destination_IP/system. . I wonder if it is possible to include the password in the URI so that the password procedure can be avoided? Thanks, Yi

3 4

[libvirt-users] editing pools
by Dave Allan 25 Jun '12

25 Jun '12

On Fedora 17 I used virsh pool-edit to change the directory of the default pool that virt-manager creates, and I noticed that my changes were discarded unless I made the changes with the pool stopped. Is that expected behavior? Dave

3 2

[libvirt-users] PCI Passthrough
by Michael Hierweck 25 Jun '12

25 Jun '12

Hi, I'm trying to passthrough legacy PCI ISDN controller to a KVM based virtual machine: <hostdev mode='subsystem' type='pci' managed='yes'> <source> <address domain='0x0000' bus='0x04' slot='0x00' function='0x0'/> </source> <address type='pci' domain='0x0000' bus='0x00' slot='0x06' function='0x0'/> </hostdev> This only worked once. (The device disappeared from the lspci autoput the host and was listed on the guest.) After shutting down the virtual maschine it refused to start again. Even rebooting the host did not help: Failed to assign irq for "hostdev0": Input/output error Perhaps you are assigning a device that shares an IRQ with another device? lspci -t -[0000:00]-+-00.0 +-19.0 +-1a.0 +-1c.0-[01]----00.0 +-1c.4-[02]----00.0 +-1c.5-[03]----00.0 +-1d.0 +-1e.0-[04]----00.0 +-1f.0 +-1f.2 \-1f.3 lspci -vv |grep IRQ Interrupt: pin A routed to IRQ 44 Interrupt: pin B routed to IRQ 22 Interrupt: pin A routed to IRQ 20 Interrupt: pin B routed to IRQ 48 Interrupt: pin D routed to IRQ 22 Interrupt: pin A routed to IRQ 16 Interrupt: pin A routed to IRQ 16 Interrupt: pin A routed to IRQ 3 Interrupt: pin A routed to IRQ 16 The last line "pin A routed to IRQ 16" belongs to my device: 04:00.0 Communication controller: Tiger Jet Network Inc. Tiger100APC ISDN chipset Subsystem: Device 0054:0001 Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx- Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=medium >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx- Interrupt: pin A routed to IRQ 16 Region 0: I/O ports at 2000 [size=256] Region 1: Memory at e2600000 (32-bit, non-prefetchable) [size=4K] Capabilities: [40] Power Management version 1 Flags: PMEClk- DSI+ D1- D2+ AuxCurrent=0mA PME(D0+,D1-,D2+,D3hot+,D3cold-) Status: D0 NoSoftRst- PME-Enable- DSel=0 DScale=0 PME- Note: After rebooting the host lscpi claims that "pin A is routed to interrupt 255" for that device. The invocation to virsh start <vm> changes that to 16. Nearly every document about PCI passthrough claims that unshared interrupts are required. But how can this be achieved? (My BIOS supports interrupt remapping for VT-d. This setting is enabled.) Thanks in advance Michael -- EDV-Serviceteam Annika & Michael Hierweck GbR Egerstraße 53, 44225 Dortmund (Germany) http://www.edv-serviceteam.net

1 0

[libvirt-users] unable to creating/list storage pools using non-root user
by mallapadi niranjan 23 Jun '12

23 Jun '12

Hi all I have a Fedora release 17 (Beefy Miracle) with libvirt versions: libvirt-0.9.11.3-1.fc17.x86_64 virt-manager-0.9.1-3.fc17.noarch I have allowed non-root user to user libvirt by allowing the user through polkit cat /etc/polkit-1/localauthority/50-local.d/cat 50-org.example-libvirt-remote-access.pkla [Remote libvirt SSH access] Identity=unix-group:virt Action=org.libvirt.unix.manage;org.libvirt.unix.monitor ResultAny=yes ResultInactive=yes ResultActive=yes After doing the above i am able to connect to virt-manager as non-root user but unable to create storage pools. [juno@reserved ~]$ id uid=1001(juno) gid=1001(juno) groups=1001(juno),1002(virt) context=staff_u:staff_r:staff_t:s0 [juno@reserved ~]$ virsh Welcome to virsh, the virtualization interactive terminal. Type: 'help' for help with commands 'quit' to quit virsh # pool-list error: Failed to reconnect to the hypervisor error: no valid connection error: Failed to connect socket to '@/home/juno/.libvirt/libvirt-sock': Connection refused virsh # list error: Failed to reconnect to the hypervisor error: no valid connection error: Failed to connect socket to '@/home/juno/.libvirt/libvirt-sock': Connection refused I have defined pool called virt-images (/virt-images) which the non-root (in this case the username is Juno) user has the read/write permissions Also tried adding the permissions to unix socket in /etc/libvirt/libvirtd.conf as below: cat /etc/libvirt/libvirtd.conf | grep -v ^$ | grep -v ^# unix_sock_group = "virt" unix_sock_ro_perms = "0777" unix_sock_rw_perms = "0770" unix_sock_dir = "/var/run/libvirt" But the unix socket are created in /var/run/libvirt and not in users home directory, So how do we make a non-root user virsh commands check the socket created in /var/run/libvirt. It always checks for the socket in user's home directory ? Any pointers on above would be helpfu. Regards Niranjan

2 6

[libvirt-users] virtio_rng in a KVM VM
by Marc Haber 22 Jun '12

22 Jun '12

Hi, I am virtualizing with KVM via libvirt and virt-manager. To have decent amounts of Entropy in the guest, I tried loading virtio_rng in a running VM. Despite the module being loaded, I didn't get any log entries, and also no /dev/hwrng. Do I need to enable the virtualized random generator in libvirt/virt-manager? If so, how do I do that? Is the support for this rather recent feature already implemented in qemu-kvm 1.0, libvirt 0.9.12 and virt-manager 0.9.1? Any hints will be appreciated. Greetings Marc -- ----------------------------------------------------------------------------- Marc Haber | "I don't trust Computers. They | Mailadresse im Header Mannheim, Germany | lose things." Winona Ryder | Fon: *49 621 31958061 Nordisch by Nature | How to make an American Quilt | Fax: *49 621 31958062

2 2

[libvirt-users] Remote virtualization with Xen, feasibility
by Jørgen Nordmoen 22 Jun '12

22 Jun '12

Hello. I'm wondering if a scenario which I'm tasked to setup would be feasible to create with libvirt. The scenario is; we want to have one computer running ns-3[1] which will simulate a MANET[2] this computer will be connected to a second computer(how is not quite clear to me yet) this second computer will be running several Xen or LXC domains where each domain is connected to a node in the ns-3 simulation. Inside each domain we will run an program which will use service-discovery to interact with all the other domains which from the point of view of a specific domain will look like wireless nodes in a network(accomplished with the ns-3 simulation, connecting the ns-3 nodes to taps[3] outside the simulation and then connecting these taps to the remote computer running libvirt and Xen or LXC). What I'm wondering is if the libvirt side of this is feasible, namely creating several domains remotely, on one or hopefully more computers, having the domains on the same computer share a storage area where we can place shared programs and where programs running inside the domains can log output to. Then after the program is done retrieve these output logs and shutdown all of the domains. When it comes to Xen, we will most likely crate a snapshot of a running Linux distribution that we can quickly start so we don't need to do a full boot process every time. Below is a very simple illustration of what we want to accomplish. Each Xen domain is connected to a specific tap device in the ns-3 server. ns-3 server libvirt server +----------------------+ +-----------------------------+ | | | +-----+ +------+ | | +-------------+ | | |Xen1| ... |XenX| | | | ns-3 | | | +-----+ +------+ | | +-------------+ | | | | tap-1 ... tap-X | | | +----------------------+ +-----------------------------+ | | | | ------------------------------------------------------------------------------------ LAN Is this possible to create with the help of libvirt? Ps. My reasoning to want to do this with libvirt is that we get support for both Xen, QEMU/KVM and LXC and libvirt has, from my first impressions, good support for Python which means quicker prototyping and better looking code. Best regards Jørgen Nordmoen [1] http://www.nsnam.org [2] https://en.wikipedia.org/wiki/MANET [3] https://en.wikipedia.org/wiki/TUN/TAP

1 0

[libvirt-users] Libvert live migration problem!
by GaoYi 22 Jun '12

22 Jun '12

Hi all, I've succeed to do live migration between two PCs. However, I found I could not connect the VM by host IP from a remote PC before migration and can only connect it from host PC. If I appended the VM with the following line at the end of the xml file: <listen type='address' address=host_IP port='PORT'> I can connect the VM via VNCVIEWER with host_IP:PORT. However, the libvirt reported: error: failed to bind the <host_IP, PORT>...It seems that the IP:port is already bound. Can someone show how to do live migration using NFS? Thanks, Yi

2 1

[libvirt-users] Cannot create macvlan devices on this platform
by Michael Hierweck 21 Jun '12

21 Jun '12

Hi, libvirt (0.9.11) refuses to start KVM based virtual machines on my system when changing the network connection from "host bridge" to "direct" (macvtap/macvlan), neither in "bridge" nor in "vepa" mode: "Cannot create macvlan devices on this platform" That's astonishing because I can easily setup working macvlan devices using the "ip" command: "ip link add link eth0 type macvlan". Would someone help me figure out the reason? Thanks in advance Michael -- EDV-Serviceteam Annika & Michael Hierweck GbR Egerstraße 53, 44225 Dortmund (Germany) http://www.edv-serviceteam.net

2 1