March 2011 - Users - libvirt List Archives

[libvirt-users] Libvirt with multipath devices and/or FC on NPIV
by C.D. 14 Mar '11

14 Mar '11

Hello, I am trying to find out a best practice for a specific scenario. First of all I would like to know what is the proper way to set up multipath, who should care about it the host or the guest. Right now I have a setup where I have one multipath which sets my host to boot from FC SAN. I have another multipathed LUN in the host which is essentially a dm which I attached to a guest, however through virtio. I added the devices through virsh with pool type of "mpath" and path through /dev/mapper/. So here is my first question: with such a setup the multipath and eventual fail-over is taken care by the host OS, right, the guest will not notice is I suppose? But what about migration. What if I decide to migrate the guest to another host. How would that work out? WIth shared directory it is easy, you just have the images in it, but what about such a setup. I can always add the aforementioned LUN, where the guest resides to another Storage Group on the SAN where the new host has access and I can make sure that the mpath device name is persistent across all hosts, but is that the right approach. Here is another question that is bugging me. I have FC HBA-s on all hosts and I would like to make a HBA visible through my guest. I stumbled upon a bug that is documented in redhat's bugzilla that attempted creation of NPIV on pci_0000_blah_blah instead on the scsi_hostX device, but I fixed that easily in the xml and although virsh doesn't seem to think I completed the nodedev-create NPIV_for_my_FC.xml the device is seen as a child in nodedev-list and what is more important the WWN are seen on the SAN switch. So this seems to be working OK, however I don't know how to attach this new shiny device that I created to a guest. Could someone give me a hint? What is the proper way to attach this new devices to a guest OS and do that persistently. Do I do that with virsh's attach-device and if so what is the proper XML format? Should I dump the XML for the newly created NPIV nodedev and try to attach that? And again with multipathing. What if I decided to create NPIV on my two FC cards in every host, then do the zoning and attach those newly created NPIV nodes to the guests? Will that produce the same effect that such a multipath does on the host? It doesn't really matter if I would use it for root drive or a shared gfs2 storage between guests, I just want to know is this possible and is it the right thing to do, or should I stick to setup in my previous paragraph? And last question: what about migration with such NPIV FC devices. I can move a guest around, but will that move my NPIV FC with it's WWN so that I can continue using my zoning (I think I just realized that this might be impossible because I'm migrating the guest, not the entire setup of libvirt, but probably I'm wrong and I'm not aware of a proper method to do it correctly?). Probably those questions are easy to answer, but please bare with me, I'm playing with libvirt in an enterprise setup only for the last couple of days and I would really like to make it right and kick some VMware ass. By the way, all my hosts are Fedora 14, with kernel 2.6.37 (rebuild of the one for Fedora 15 from koji, because there was a small glitch with Qlogic's driver in the stock 2.6.35). Most of my guests will be SL6,CentOS5/6(when it arrives) and a couple of windows XP guests (but I'm not really concerned with that). Thanks for the support in advance

1 0

[libvirt-users] Libvirt with multipath devices and/or FC on NPIV
by C.D. 14 Mar '11

14 Mar '11

Hello, I am trying to find out a best practice for a specific scenario. First of all I would like to know what is the proper way to set up multipath, who should care about it the host or the guest. Right now I have a setup where I have one multipath which sets my host to boot from FC SAN. I have another multipathed LUN in the host which is essentially a dm which I attached to a guest, however through virtio. I added the devices through virsh with pool type of "mpath" and path through /dev/mapper/. So here is my first question: with such a setup the multipath and eventual fail-over is taken care by the host OS, right, the guest will not notice is I suppose? But what about migration. What if I decide to migrate the guest to another host. How would that work out? WIth shared directory it is easy, you just have the images in it, but what about such a setup. I can always add the aforementioned LUN, where the guest resides to another Storage Group on the SAN where the new host has access and I can make sure that the mpath device name is persistent across all hosts, but is that the right approach. Here is another question that is bugging me. I have FC HBA-s on all hosts and I would like to make a HBA visible through my guest. I stumbled upon a bug that is documented in redhat's bugzilla that attempted creation of NPIV on pci_0000_blah_blah instead on the scsi_hostX device, but I fixed that easily in the xml and although virsh doesn't seem to think I completed the nodedev-create NPIV_for_my_FC.xml the device is seen as a child in nodedev-list and what is more important the WWN are seen on the SAN switch. So this seems to be working OK, however I don't know how to attach this new shiny device that I created to a guest. Could someone give me a hint? What is the proper way to attach this new devices to a guest OS and do that persistently. Do I do that with virsh's attach-device and if so what is the proper XML format? Should I dump the XML for the newly created NPIV nodedev and try to attach that? And again with multipathing. What if I decided to create NPIV on my two FC cards in every host, then do the zoning and attach those newly created NPIV nodes to the guests? Will that produce the same effect that such a multipath does on the host? It doesn't really matter if I would use it for root drive or a shared gfs2 storage between guests, I just want to know is this possible and is it the right thing to do, or should I stick to setup in my previous paragraph? And last question: what about migration with such NPIV FC devices. I can move a guest around, but will that move my NPIV FC with it's WWN so that I can continue using my zoning (I think I just realized that this might be impossible because I'm migrating the guest, not the entire setup of libvirt, but probably I'm wrong and I'm not aware of a proper method to do it correctly?). Probably those questions are easy to answer, but please bare with me, I'm playing with libvirt in an enterprise setup only for the last couple of days and I would really like to make it right and kick some VMware ass. By the way, all my hosts are Fedora 14, with kernel 2.6.37 (rebuild of the one for Fedora 15 from koji, because there was a small glitch with Qlogic's driver in the stock 2.6.35). Most of my guests will be SL6,CentOS5/6(when it arrives) and a couple of windows XP guests (but I'm not really concerned with that). Thanks for the support in advance P.S. please keep me in CC, as I'm not on the list. Thank you

1 0

[libvirt-users] Issue creating vm channel
by SAURAV LAHIRI 12 Mar '11

12 Mar '11

Hello, I am trying to create a lucid guest VM with a vmchannel between the guest and the hypervisor. Details below: KERNEL: # uname -a Linux saurav-desktop 2.6.38-2-generic #29~lucid1-Ubuntu SMP Mon Feb 7 13:35:14 UTC 2011 i686 GNU/Linux QEMU: # /usr/bin/qemu --version QEMU emulator version 0.13.0, Copyright (c) 2003-2008 Fabrice Bellard LIBVIRT: # /usr/sbin/libvirtd --version /usr/local/sbin/libvirtd (libvirt) 0.8.8 Below is a snippet of my libvirt xml file. =============================================================== <domain type='kvm' id='16'> <name>FinalTest</name> <uuid>212bc1f1-b4f7-1ed3-7dc3-2b4b82ab68d1</uuid> <memory>1048576</memory> <currentMemory>1048576</currentMemory> <vcpu>1</vcpu> <os> <type arch='x86_64' machine='pc-0.12'>hvm</type> <boot dev='hd'/> </os> <features> <acpi/> <apic/> <pae/> </features> <clock offset='utc'/> <on_poweroff>destroy</on_poweroff> <on_reboot>destroy</on_reboot> <on_crash>destroy</on_crash> <devices> <emulator>/usr/bin/kvm</emulator> <controller type='virtio-serial' index='0' max_ports='15' vectors='4'/> <channel type='pty'> <target type='virtio' name='org.linux-kvm.port.1'/> <address type='virtio-serial' controller='0' bus='0'/> </channel> <disk type='file' device='disk'> <driver name='qemu' type='qcow2'/> <source file='/home/user1/linux_installed/image.cow2'/> <target dev='hda' bus='ide'/> <alias name='ide-disk0'/> <address type='drive' controller='0' bus='0' unit='0'/> </disk> ================================================================ When i submit the xml file to virsh I get the following error: =============================================================== # virsh create vmTemplete-1.xml error: Failed to create domain from vmTemplete-1.xml error: internal error Process exited while reading console log output: char device redirected to /dev/pts/3 char device redirected to /dev/pts/4 open /dev/kvm: No such file or directory Could not initialize KVM, will disable KVM support kvm: -device virtserialport,bus=virtio-serial0.0,nr=1,chardev=charchannel0,id=channel0,name=org.linux-kvm.port.1: virtio-serial-bus: Out-of-range port id specified, max. allowed: 0 kvm: -device virtserialport,bus=virtio-serial0.0,nr=1,chardev=charchannel0,id=channel0,name=org.linux-kvm.port.1: Device 'virtserialport' could not be initialized =========================================================== What could be the problem ? Could any one please point to the correct direction with regard to this. Regards sl

1 1

[libvirt-users] dnsmasq not started when no dhcp enabled ?
by jbd 12 Mar '11

12 Mar '11

Hello everybody, i've defined a simple network, with no dhcp. I'd like to use dnsmasq only as a dns server. $ virsh net-dumpxml basicswitch <network> <name>basicswitch</name> <uuid>60f491d2-d6c4-6b57-8a50-081cace8dedc</uuid> <forward mode='nat'/> <bridge name='virbr1' stp='on' delay='0' /> <ip address='192.168.100.1' netmask='255.255.255.0'> </ip> </network> I start it : $ virsh net-start basicswitch Network basicswitch started $ virsh net-list Name State Autostart ----------------------------------------- basicswitch active no But i see no dnsmasq process : $ ps fax|grep dnsmas[q] Here the output of "virsh version" : $ virsh version Compiled against library: libvir 0.8.8 Using library: libvir 0.8.8 Using API: QEMU 0.8.8 Running hypervisor: QEMU 0.12.5 What is funny is that the dnsmasq process is created on another box with different version : $ virsh version Compiled against library: libvir 0.8.3 Using library: libvir 0.8.3 Using API: QEMU 0.8.3 Running hypervisor: QEMU 0.12.5 $ ps fax|grep dnsmas[q] 7104 ? S 0:00 dnsmasq --strict-order --bind-interfaces --pid-file=/var/run/libvirt/network/basicswitch.pid --conf-file= --listen-address 192.168.100.1 --except-interface lo Any advice on it ? Regards, Jean-Baptiste

2 2

[libvirt-users] libvirt configuration problem
by Onkar Mahajan 12 Mar '11

12 Mar '11

Hi , I am getting errors while configuring libvirt to compile it from source. There is a error in libnl checking for UDEV... no checking whether to compile with macvtap support... yes checking whether to compile with virtual port support... no checking for LIBNL... no configure: error: libnl-devel >= 1.1 is required for macvtap support I have already installed the libnl library in /lib/ [root@localhost lib]# ls libnl* libnl.a libnl-cli.so.2 libnl-genl.so libnl-nf.a libnl-nf.so.2.0.0 libnl-route.so.2 libnl.so.2.0.0 libnl-cli.a libnl-cli.so.2.0.0 libnl-genl.so.2 libnl-nf.la libnl-route.a libnl-route.so.2.0.0 libnl-cli.la libnl-genl.a libnl-genl.so.2.0.0 libnl-nf.so libnl-route.la libnl.so libnl-cli.so libnl-genl.la libnl.la libnl-nf.so.2 libnl-route.so libnl.so.2 Please help me !! Regards, Onkar

2 1

[libvirt-users] libvirt/kvm/qemu: pointopoint routed setup?
by Andreas Jellinghaus 12 Mar '11

12 Mar '11

Hi, can anyone give an example for a pointopoint routed setup? i.e. each virtual machine has one ip with pointopoint config to the host machine, on a private interface - thus the virtual machines can only talk to the host, which routes them and can use normal iptables for filtering all traffic. with xen this was a simple script doing ifconfig ${vif} ${main_ip} netmask 255.255.255.255 up ip route ${ipcmd} ${addr} dev ${vif} src ${main_ip} and an optional echo 1 >/proc/sys/net/ipv4/conf/${vif}/proxy_arp has someone implemented something like this with libvirt/kvm setup? can you give some pointers how to do this? I don't need high speed communication between the virtual machines, and the option to filter all traffic between them (without using the bridge netfilter tables) would be nice. Or is there a reason not to use such a setup, and a better option to implement this? Thanks for your help and best regards, Andreas Jellinghaus

1 0

[libvirt-users] Migrating existing debian system to KVM/libvirt
by Randall Webstuff 12 Mar '11

12 Mar '11

I have a dying machine running debian etch that acts as a web/mail server that I would like to virtualize using debian/kvm/libvirt. Is it as simple as removing the hard-disk (multiple ext2/ext3 partitions) from the dying machine into my newer server and then writing a custom xml config file? Is there a tutorial on doing this somewhere? Thanks. Randy

2 1

[libvirt-users] client certificate path hard coded?
by Anthony Goddard 12 Mar '11

12 Mar '11

Hi all, I'm trying to figure out how to get my desktop talking to two libvirt hosts using qemu+tls and I've read that virsh relies on hard coded paths to the certificates.. which seems to be true. Is there a way to tell virsh to use a different path to a certificate, or another way people solve this presently? Cheers, Ant

2 2

[libvirt-users] libvirt and phyp driver (Power Hypervisor guests)
by Boninsegna, Werner (Werner, ES-Best-Shore-Services-China-SH) 11 Mar '11

11 Mar '11

Hi all, I would like to ask about the status of the phyp driver to manage Power Hypervisor guests. According the documentation [1] the driver is supported, however it does not show up on the main driver page [2]. Has somebody already had experience using this driver in a productive or testing environment? [1] http://libvirt.org/guide/html/Application_Development_Guide-Connections-URI… [2] http://libvirt.org/drivers.html Kind regards, Werner --- Werner Boninsegna HP Enterprise Services

1 0

[libvirt-users] unable to connect to libvirtd at '*': No route to host
by Christoph Raible 10 Mar '11

10 Mar '11

Hi @all, I have the following systems: AMD Processor Scientific Linux 6.0 Kernelversion: 2.6.32-71.18.2.el6.x86_64 KVM / Libvirt out of the repository I've setup my TLS-Certificates with the following howto /http://wiki.libvirt.org/page/TLSSetup/ Now I want to connect the servers with /virsh -c qemu+tls://nebula3/system/(nebula3 is the hostname) but the following error message always is displayed: /error: unable to connect to libvirtd at 'nebula3': No route to host error: failed to connect to the hypervisor/ SELinux is disabled and the servers are in the same IP-Range (ping works fine) Passwordless ssh-login also works fine! If I connect to a server with qemu+ssh://nebula3/system all wroks fine... Installation with virt-manager an bridged networkinterface also works fine... Did anyone of you have an idea where my problem is? Thanks for Help :) Regards, Christoph -- Vorstand/Board of Management: Dr. Bernd Finkbeiner, Dr. Roland Niemeier, Dr. Arno Steitz, Dr. Ingrid Zech Vorsitzender des Aufsichtsrats/ Chairman of the Supervisory Board: Michel Lepert Sitz/Registered Office: Tuebingen Registergericht/Registration Court: Stuttgart Registernummer/Commercial Register No.: HRB 382196

2 3