- Users - libvirt List Archives

[libvirt-users] How to configure MacVtap passthrough mode to SR-IOV VF?
by opendaylight 20 Jan '14

20 Jan '14

Hi guys. These days I'm doing research on SR-IOV & Live migration. As we all know there is big problem that SR-IOV & Live migration can not exist at the same time. I heard that KVM + SRIOV + MacVtap can solve this problem. So I want to try. My environment: Host: Dell R610, OS: RHEL 6.4 ( kernel 2.6.32) NIC: intel 82599 I follow a document from intel guy, it said that I should write xml like below: ============================ <network> <name>macvtap_passthrough’</name> <forward mode=’passthrough> <interface dev=’vf0’ /> <interface dev=’vf1’ /> .. .. </forward> </network> ============================ I guess here the vf0 & vf1 should be the VFs of Intel 82599. What make me confused is that we know we can not see the vf 0 & vf 1 directly from the host server with "ifconfig", that is to say, vf 0 & vf1 are not a real physical interface. I try #: virsh net-define macvtap_passthrough.xml #: virsh net-start macvtap_passthrough When I try to configure macvtap_passthrough for a VNIC of a VM, the virt-manager told : "Can't get vf 0, no such a device". When I try from virt-manager: add hardware--->network--->host device (macvtap_passthrough:pass_through network), I got error like : "Error adding device: xmlParseDoc() failed". I guess I can not write like this " <interface dev=’vf0’ />" in the xml. I try to change as below, but the result is same. ============================ <network> <name>macvtap_passthrough’</name> <forward mode=’passthrough> <pf dev=’p2p1’ /> // p2p1 is intel sriov physical nic </forward> </network> ============================ I don't know how to write correctly. Please help me. You can refer to intel document as below. Many thanks. ==========document from intel======================== Linux/KVM VM Live Migration (SRIOV And MacVtap) By Waseem Ahmad (waseem.ahmad(a)intel.com In this scenario we are using 3 machines:  Server 1: DNS/NFS – nfs.vtt.priv  Server 2: Hv1  Server3: Hv2 HV1 and HV2 are Linux/KVM machines. We will get to them in a minute however we first must address kvm and nfs. NFS:  Create a storage area, where both HV1, and HV2 can access it. There are several methods available for this (FCOE/ISCSI/NFS). For this write-up use nfs. Configure NFS:  Create a directory on nfs.vtt.priv where you want your storage to be. In this case used /home/vmstorage  Edit /etc/exports and add the following /home/vmstorage 172.0.0.0/1(rw,no_root_squash,sync)  Now to /etc/sysconfig/nfs Uncomment RPCNFSDARGS=”-N 4” This will disable nfs v4. If you don’t do this you will have issues with not being able to access the share from within VirtManager.  Add all three machines ip addresses to each machines /hosts file. MIGRATION WILL NOT WORK WITHOUT FULLY QUALIFIED DOMAIN NAMES. KVM: On both HV1, and HV2 servers:  Edit /etc/selinux/config SELINUX=disabled  Edit /etc/libvirt/qemu.conf Change security_driver=none  On HV1 and HV2 start Virtual Machine Manager  Double click on localhost(QEMU)  Then click on the storage tab at the top of the window that pops up  Down in the left hand corner is a box with a + sign in it, click on that. A new window will appear entitled Add a New Storage Pool  In the name box type vmstorage, then click on the type box and select netfs: Network Exported Directory, now click next.  You will see the last step of the network Storage Pool Dialog. The first option is the target path. This is the path where we will mount our storage on the local server. I have chosen to leave this alone.  The next option is format, leave this set on auto: Host name: nfs.vtt.priv Source path: /home/vmstorage  Click on finish  Repeat the above steps on HV2 server Create vms  On HV1 server go back to the connection details screen, (this is the one that showed up when you double clicked on localhost (qemu), and click on the storage tab again.  Click on vmstorage then click on new volume at the bottom.  A new dialog will appear entitled add a storage volume.  In the Name box type vm1  In the Max Capacity box type 20000  And do the same in the allocation box then click finish.  Now you can close the connection details box by click on the x in the corner.  Now click on the terminal in the corner, right underneath file, and type the name of our vm in the box that is entitled Name, vm1 choose how your installation media, probably local install media, and click forward. Click on use cdrom or dvd, and place a rh6.2 dvd in the dvd player on HV1. Select Linux, for the OS type, and Red Hat Enterprise Linux 6 for the version. Memory I chose to leave this at its default of 1024, and assigned 1 cpu to the guest. Click forward, select “select managed or other existing storage” and click the browse button. Click on vmstoarge, and select vm1.img then click forward. Then click on finish.  We will configure network after we make sure migration between the two servers works properly.  Now go ahead and install the operating system as you would normally. Create networks Create a file that looks like the following < there is no support for adding a macvtap interface from the gui as of yet, this is the only manual step in the process. Create a file named macvtap_passthrough.xml with the following contents. <network> <name>macvtap_passthrough’</name> <forward mode=’passthrough> <interface dev=’vf0’ /> <interface dev=’vf1’ /> .. .. </forward> </network> <network> <name>’macvtap_bridge’</name> <forward mode=’bridge’> <interface dev=’p3p1’/> </forward> </network> Save it and run the following commands:  virsh net-define macvtap_passthrough.xml  virsh net-start macvtap_passthrough Make sure all of your virtual interfaces that you used in the xml file are up. for i in $(ifconfig –a | awk ‘/eth/ {print $1}’); do ifconfig $i up; done Then double click on your vm and click on the big blue i On the next screen click on add hardware, then on network, then select Virtual network “macvtap_passthrough” Then click on finish. Start your vm and make sure that the macvtap was created on the host by doing ip link | grep ‘macvtap’ In the vm configure the ip information for the virtio adapter. In the virtual machine manager click on file, add connection. Then check the connect to remote host fill in the username and hostname, then click on connect Right click on your VM and select Migrate, select the host you want to migrate the machine to, then click on advanced options, check the address box, and type the ip address of the machine you want to migrate to, and click the migrate button.

2 1

[libvirt-users] How to update filterref of a vm on the fly?
by Gao Yongwei 17 Jan '14

17 Jan '14

Hello, I defined a vm with filterref like: <filterref filter='clean-traffic'> <parameter name='IP' value='192.168.1.161'/> </filterref> and now I need to add another IP parameter for this vm,is there any way to achieve this? thanks.

4 4

[libvirt-users] blockcopy, userspace iSCSI support?
by Scott Sullivan 17 Jan '14

17 Jan '14

Right now, on a virsh blockcopy, I know you can do something like this: # Connect DEST target iscsiadm -m node -p ${DESTINATION}:3260 -T ${VOLNAME} -o new iscsiadm -m node -p ${DESTINATION}:3260 -T ${VOLNAME} --login # Copy to connected iSCSI target virsh blockcopy ${DOMAIN} vda /dev/sdc --raw --bandwidth 300 However I have libiscsi compiled into my QEMU. So I can do this with the monitor directly (and avoid the need to call out to external iscsiadm): virsh qemu-monitor-command ${DOMAIN} '{"execute":"drive-mirror", "arguments": { "device": "drive-virtio-disk0", "target": "iscsi://${TARGET_IP}:3260/${DOMAIN}/1", "mode": "existing", "sync": "full", "on-source-error": "stop", "on-target-error": "stop" } }' Is there a way to use the libiscsi compiled into my QEMU with virsh blockcopy command? I haven't been able to find any examples of using blockcopy with iSCSI compiled into QEMU.

1 0

[libvirt-users] Ceph RBD locking for libvirt-managed LXC (someday) live migrations
by Joshua Dotson 16 Jan '14

16 Jan '14

Hi, I'm trying to build an active/active virtualization cluster using a Ceph RBD as backing for each libvirt-managed LXC. I know live migration for LXC isn't yet possible, but I'd like to build my infrastructure as if it were. That is, I would like to be sure proper locking is in place for live migrations to someday take place. In other words, I'm building things as if I were using KVM and live migration via libvirt. I've been looking at corosync, pacemaker, virtlock, sanlock, gfs2, ocfs2, glusterfs, cephfs, ceph RBD and other solutions. I admit that I'm quite confused. If oVirt, with its embedded GlusterFS and its planned self-hosted engine option, supported LXC, I'd use that. However the stars have not yet aligned for that. It seems that the most elegant and scalable approach may be to utilize Ceph's RBD with its native locking mechanism plus corosync and pacemaker for fencing, for a number of reasons out of scope for this email. *My question now is in regards to proper locking. Please see the following links. The libvirt hook looks good, but is there any expectation that this arrangement will become a patch to libvirt itself, as is suggested by the second link?* http://www.wogri.at/en/linux/ceph-libvirt-locking/ http://lists.ceph.com/pipermail/ceph-users-ceph.com/2013-August/003887.html Can anyone guide me on how to theoretically build a very "lock" safe 5-node active-active KVM cluster atop Ceph RBD? Must I use sanlock with its NFS or GFS2 with its performance bottlenecks? Does your answer work for LXC (sans the current state of live migration)? Thanks, Joshua -- Joshua Dotson Founder, Wrale Ltd

2 2

[libvirt-users] Live migration finish threshold
by Joaquim Barrera 16 Jan '14

16 Jan '14

Hello everyone, Can somebody point where I can find the code where libvirt makes the decision to complete a live migration? I mean, at some point syncronising the VM state, it has to decide that the delta left to be migrate is low enough to achieve downtime 0, so libvirt finishes the migration. This "low enough" must be defined somewhere in the code, but I am unable to find it. Thank you very much.

2 1

[libvirt-users] vnc port/listen address ignored when setting machine?
by hubert depesz lubaczewski 16 Jan '14

16 Jan '14

Hi, First of all, I hope it's not a big problem - I'm running on Debian, not Redhat. To my problem: I'm starting to learn virtualization, libvirt, and decided to create some test machine. I did it with: virt-install --name debian-test \ --os-type=linux \ --os-variant=debianwheezy \ --cdrom /media/media/software/iso/debian-testing-amd64-netinst-2014-01-16.iso \ --graphics vnc,listen=0.0.0.0,port=20001 \ --disk pool=default,format=raw,size=20 \ --ram 2048 \ --vcpus=2 \ --network bridge=virbr0 \ --hvm \ --virt-type=kvm Machine starts, but domdisplay shows: =# virsh domdisplay debian-test vnc://localhost:14101 When I tried setting it with --grpahics=vnc,...port=40001, I got it to listen on port 34101 (which is weird anyway, but at least I can change it). But whatever I do - I can't make it to listen on 0.0.0.0. I know I can use ssh port forwarding to get to this VNC, but I would very much prefer to be able to set VNC listen to 0.0.0.0 for at least some of my test domains. What am I doing wrong? Best regards, depesz -- The best thing about modern society is how easy it is to avoid contact with it. http://depesz.com/

2 3

[libvirt-users] Does libvirt lxc driver support "cpuset" attribute?
by WANG Cheng D 16 Jan '14

16 Jan '14

Dear all I allocate only one vcpu for the container by the following statement, that is, I want to pin the vcpu to physical core "2". <vcpu placement='static' cpuset="2" >1</vcpu> My host has 4 physical cores. Before test, all the 4 cores are idle. After I run 4 processes in the container, I found all the 4 cores in the host are 100% used. That is, the container is pinned to all the available physical CPUs. This is not what I want. My libvirt version is 1.0.3. Although "vcpupin" element also can be used to pin vcpu, according to http://libvirt.org/formatdomain.html , "vcpupin" element is not supported by lxc driver. I wonder if it is the older version of libvirt that causes the problem? Thank you in advance. Cheng Wang

2 2

[libvirt-users] Fake Network Interface
by Andrew Martin 15 Jan '14

15 Jan '14

Hello, Is there a supported method for creating a fake network interface in a VM's configuration file? I was using the below construct, however it is no longer working for me in recent versions of libvirt (libvirt 1.0.2 with qemu-kvm 1.4.0). Is there a different, preferred method for creating a fake virtual interface on the guest which does not exist on the host? <interface type='network'> <mac address='aa:aa:aa:aa:aa:aa'/> <source network='fake'/> <address type='pci' domain='0x0000' bus='0x00' slot='0x03' function='0x0'/> </interface> Thanks, Andrew

2 3

[libvirt-users] virDBusCallMethod:1173 : Launch helper exited with unknown return code
by Karoline Haus 15 Jan '14

15 Jan '14

Good morning, I'm using libvirtd on Gentoo. This is libvirt version: 1.1.3.1 I have trouble starting a VM using virsh start $vm. I do this as root, because as non-root user it did not work at all (especially it failed attaching to the networks). So, when I run the command (with sudo), I get the following error in libvirtd.log: 2014-01-15 07:51:00.423+0000: 16158: warning : qemuDomainObjTaint:1573 : Domain id=5 name='vader' uuid=f5b8c05b-9c7a-3211-49b9-2bd635f7e2aa is tainted: high-privileges 2014-01-15 07:51:00.428+0000: 16158: error : virDBusCallMethod:1173 : Launch helper exited with unknown return code 1 At the same time I get an error in /var/log/messages which seems related: Jan 15 07:51:00 dbus[15845]: [system] Activating service name='org.freedesktop.machine1' (using servicehelper) Jan 15 07:51:00 dbus[15845]: [system] Activated service 'org.freedesktop.machine1' failed: Launch helper exited with unknown return code 1 Anyone ever seen this issue? I have no idea where to look for errors because the message don't really tell me much. I have tried to execute the qemu-kvm command on the command line directly and that worked immediately. So the problem must be in libvirt. Thanks for any pointers.

3 3

[libvirt-users] Best practice for custom iptables rules
by ZeroUno 15 Jan '14

15 Jan '14

Hi, I'm using libvirt to manage some VMs on a CentOS host, and I need some custom iptables rules to always be in place for some communications to happen, e.g. between the VMs and the outside world in both directions. Some of these rules need to be at the top of the iptables chain, otherwise the default rules added by libvirt would block the communications I need. So I cannot just add the rules in /etc/sysconfig/iptables, because libvirt adds its own rules _before_ the rules contained in this config file. I was looking at filters, but maybe not every rule can be made into a filter? Specifically, I need a rule for the POSTROUTING chain in the "nat" table. Can it be added through filters? Also, regarding the "iptables restart problem" described in the last paragraph at <http://libvirt.org/firewall.html>, is there really no acceptable way to make libvirt add its rules back automatically upon iptables/network restart? Thanks for any info. Marco -- 01

4 12