- Users - libvirt List Archives

[libvirt-users] High availability for guests
by Renich Bon Ciric 29 Sep '12

29 Sep '12

Hello, What is the recommended way to offer high availability to Guests? For example, let host1, host2 and host3 be libvirt and KVM/Qemu enabled. If I start a guest in host1, how can I guarantee that it will stay online if host1 goes down? GlusterFS, for example, would take care of storage; but what about CPU and RAM? How can this be accomplished? Thank you for any feedback or comment in adavance. -- It's hard to be free... but I love to struggle. Love isn't asked for; it's just given. Respect isn't asked for; it's earned! Renich Bon Ciric http://www.woralelandia.com/ http://www.introbella.com/

2 3

[libvirt-users] Inconsistent iptables forwarding rules for virtual networks?
by Marwan Tanager 27 Sep '12

27 Sep '12

Hi everyone. Those are the iptables forwarding rules associated with the two virtual networks on my machine: ----------------------------------------------------------------------------------------- -A FORWARD -d 192.168.100.0/24 -o virbr1 -m state --state RELATED,ESTABLISHED -j ACCEPT -A FORWARD -s 192.168.100.0/24 -i virbr1 -j ACCEPT -A FORWARD -i virbr1 -o virbr1 -j ACCEPT -A FORWARD -o virbr1 -j REJECT --reject-with icmp-port-unreachable -A FORWARD -i virbr1 -j REJECT --reject-with icmp-port-unreachable -A FORWARD -d 192.168.122.0/24 -o virbr0 -m state --state RELATED,ESTABLISHED -j ACCEPT -A FORWARD -s 192.168.122.0/24 -i virbr0 -j ACCEPT -A FORWARD -i virbr0 -o virbr0 -j ACCEPT -A FORWARD -o virbr0 -j REJECT --reject-with icmp-port-unreachable -A FORWARD -i virbr0 -j REJECT --reject-with icmp-port-unreachable ----------------------------------------------------------------------------------------- Using this setup I can ping from a guest on 192.168.100.0/24 to a guest on 192.168.122.0/24, however the reverse is not possible (I get 'port unreachable' icmp messages). This is caused by the fourth rule, which rejects packets forwarded to virbr1. Moving the fourth rule to the end, obviously eliminates this problem. So, is this default setup intentional for some reason, or should the rules just be rearranged by libvirt, so that any rule on the FORWARD chain which has a REJECT target (which reject forwarding "to" some virtual interface) would go to the end of the chain, and thus, allowing the virtual networks to reach each other by default? Thanks. Marwan

2 2

[libvirt-users] minor swap issue ....
by William A. Mahaffey III 25 Sep '12

25 Sep '12

.... I have a CentOS 5.n VM running on a Fedora 14 server/host, installed using virt-manager, pretty plain vanilla. I use it to compile binaries to run under RHEL/CentOS 5.n OS. I occasionally notice that when the VM gets paged out by the server, it takes several minutes to get it back in :-/ (see below). On the host, logged in to a shell through a terminal window, this A.M.: [wam@Q6600, test, 10:09:08am] 3384 % /sbin/swapon -s ; free -m ; uname -a ; /sbin/hwclock -r ; date Filename Type Size Used Priority /dev/md2 partition 16773116 1967524 -1 total used free shared buffers cached Mem: 7993 7932 61 0 11 7176 -/+ buffers/cache: 744 7249 Swap: 16379 1921 14458 Linux Q6600 2.6.35.14-106.fc14.x86_64 #1 SMP Wed Nov 23 13:07:52 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux Tue Sep 25 10:09:11 2012 -0.890900 seconds Tue Sep 25 10:09:11 CDT 2012 [wam@Q6600, test, 10:09:11am] 3385 % /sbin/swapon -s ; free -m ; uname -a ; /sbin/hwclock -r ; date Filename Type Size Used Priority /dev/md2 partition 16773116 1415136 -1 total used free shared buffers cached Mem: 7993 7939 54 0 12 6358 -/+ buffers/cache: 1568 6425 Swap: 16379 1381 14998 Linux Q6600 2.6.35.14-106.fc14.x86_64 #1 SMP Wed Nov 23 13:07:52 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux Tue Sep 25 10:14:09 2012 -0.281487 seconds Tue Sep 25 10:14:09 CDT 2012 [wam@Q6600, test, 10:14:09am] 3385 % on the VM, logged in through a different terminal window: [wam@centos-5, CFD, 10:09:14am] 6507 % [wam@centos-5, CFD, 10:09:14am] 6507 % ( cd ../Utils/ ; make extra ) ; cd . make[1]: Entering directory `/home/wam/dev/CFD/Utils' make[2]: Entering directory `/home/wam/dev/CFD/Utils' /home/wam/dev/CFD/Utils/../lib/R4/SSE2/libutils.a up to date. make[2]: Leaving directory `/home/wam/dev/CFD/Utils' make[2]: Entering directory `/home/wam/dev/CFD/Utils' /home/wam/dev/CFD/Utils/../lib/R8/SSE2/libutils.a up to date. make[2]: Leaving directory `/home/wam/dev/CFD/Utils' make[1]: Leaving directory `/home/wam/dev/CFD/Utils' cd /net/cube/home/wam/V8/Cnx/test/junk/cart/unstaggered/bfc/utils; icc -c -DNDEBUG -DUNDER_SCORE_SYS -DLOSE_GAMMAL -gcc-version=340 -gcc-name=gcc34 -fabi-version=2 -I /net/cube/home/wam/V8/Cnx/test/junk/cart/unstaggered/bfc/include -wd138,167,186,266,279,880,1418,1419,2330,2331 -O3 -ip -opt-prefetch -parallel -opt-multi-version-aggressive -xSSE2 -DP64_BIT test.cpp cd /net/cube/home/wam/V8/Cnx/test/junk/cart/unstaggered/bfc/utils; icc -o /home/wam/bin/checkout -parallel -mkl=parallel -Wl,-s test.o -L /home/wam/dev/CFD/Utils/../lib/R4/SSE2 -Wl,--start-group -lutils -lMemIO -Wl,--end-group && \rm -f test.o Done with extra. [wam@centos-5, CFD, 10:12:36am] 6508 % i.e. it took 3+ min. to get back to the prompt for this command (usually about 5 sec.) .... This isn't a show-stopped by any means, but it is irritating. From the timestamps & data from the host & guest, apparently all of that was used up swapping the guest back in. Is there any way to either prioritize the VM to not get swapped out, or preferentially swapped back in :-) ? TIA for any pointers .... -- William A. Mahaffey III ---------------------------------------------------------------------- "The M1 Garand is without doubt the finest implement of war ever devised by man." -- Gen. George S. Patton Jr.

3 4

[libvirt-users] Xen Paravirtualized Guest and drive order
by Cyril Peponnet 25 Sep '12

25 Sep '12

Hello, When defining a paravirtulized xen guest and use pygrub as a loader, it will seek into the first drive defined in order to fetch the kernel and boot the machine. As the boot tag is not used, and boot order seems to have no effect, I try to define the disk I want the kernel been seeked as first in my xml definition. Unfortunately, after defining the vm, the order is not kept. As I have seen, the order is made by alphabetical order. I mean if I define a disk as xvda and a cdrom as xvdc, the first drive will be the disk xvda. To be able to make the cdrom first I need to define xvda as a target for cdrom and xvdb or xvdc ... for the disk. It is not really usefull... Is there any way to keep the order of drive as defined in the xml config file ? Thanks

1 0

[libvirt-users] Quadratic function-like call delay
by Andrey Korolyov 25 Sep '12

25 Sep '12

Hi, There is a quite annoying thing, don`t sure if I can call it a bug. How to reproduce: - start a bunch of VMs, - stop libvirt, - start libvirt and immediately issue any call, say, 'virsh version', - delay until call completion may be fitted nearly as quadratic function from number of running VMs, qemu-kvm in my case (see link below). After this delay passed, any calls executed without slowdown, so problem is only a 'dead gap' after restarting service. http://xdel.ru/downloads/libvirt.png

4 6

[libvirt-users] VLANs with Open vSwitch in libvirt 0.10.1
by Scott Lowe 23 Sep '12

23 Sep '12

Hello all, I'm trying to take advantage of libvirt's VLAN and OVS support in the 0.10.1 release and consistently getting an "error: internal error corrupted profileid string" when starting a guest. I've defined a network using virsh; the configuration looks like this: <network> <name>ovs-network</name> <uuid>40c32a20-9eed-238c-895c-b9af7eaf72f1</uuid> <forward mode='bridge'/> <bridge name='ovsbr0' /> <virtualport type='openvswitch'/> <portgroup name='green' default='yes'> <virtualport> <parameters profileid='green-profile'/> </virtualport> </portgroup> <portgroup name='blue'> <vlan> <tag id='2'/> </vlan> <virtualport> <parameters profileid='blue-profile'/> </virtualport> </portgroup> </network> This network is listed when I run "virsh net-list --all" and is listed as active. In a guest, the network is defined like this: <interface type='network'> <mac address='52:54:00:a8:bf:8f'/> <source network='ovs-network'/> <virtualport type='openvswitch'> <parameters interfaceid='2913d936-6771-4983-1cdf-753c45ac24d2' profileid='green-profile'/> </virtualport> <address type='pci' domain='0x0000' bus='0x00' slot='0x04' function='0x0'/> </interface> This is running on Ubuntu 12.04.1 (and QEMU-KVM) with manually compiled libvirt 0.10.1 and OVS version 1.4.0+build0. For troubleshooting, I've undefined/re-define both the network and the guest, and restarted libvirtd in between. I've also tried different network names, different portgroup names, and different profileid's. I'm sure I'm missing something obvious but I can't seem to find it--any suggestions? Any and all help is appreciated. -- Scott

1 0

[libvirt-users] prefer_msi=on in xml file
by klak2 23 Sep '12

23 Sep '12

Hello, here is a problem with KVM passthrough of TechnoTrend S2-6400. Bug report is done. The last answer of the debian maintainer: --> Please check with lspci -vv whenever your device is MSI-capable, --> and if yes, try to add prefer_msi=on to the pci-assign device --> argument -- like, from your original report, --> --> -device pci- assign,host=08:00.0,id=hostdev0,configfd=23,bus=pci.0,addr=0xa,prefer_msi=on --> --> I don't know how to add this to libvirt however, maybe they have --> some knobs for that nowadays. and now my question: How do I add the parameter "prefer_msi=on" to the xml file? Thank and greets klak2

1 0

[libvirt-users] sharing read-only LOCAL disks among KVM guests
by Amos Shapira 23 Sep '12

23 Sep '12

Hello, Is it possible to share a single host's Logical Volume among multiple local KVM guests which mount it read-only? I'm asking this because I have an idea to run multiple idential KVM guests (they all have exactly the same software installed on them), booting them from a shared local Logical Volume read-only root file system, or alternatively let them share the bulk of the software (/usr, /opt, /lib) from a common KVM host Logical volume. Is this possible? All my searches so far failed to turn up anything like this. Would it be possible using qcow2 instead of raw LV? If so - would it be worth the performance hit of switching from LV to qcow2? The motivation for this is two-fold: 1. Speed-up - I suspect that sharing the same disk block among all KVM agents (we currently have 14 KVM guests on each server) should lead to better cache utilisation and reduce disk access. Also the cost of maintaining the mutiple identical servers and keeping them up to date could be eating into our server's resource usage. 2. Maintenance - if I get to share the common disk image among all of them then I only need to install/update it once instead of once for each KVM guest (we already use Puppet but it doesn't mean it comes for free). Thanks, --Amos

3 7

[libvirt-users] picking a NUMA cell for pinning using virsh freecell
by Edoardo Comar 22 Sep '12

22 Sep '12

Hi I'd want to pin the vcpu of a guest to a pcpu. the docs clearly say https://access.redhat.com/knowledge/docs/en-US/Red_Hat_Enterprise_Linux/6/h… "Locking a guest to a particular NUMA node offers no benefit if that node does not have sufficient free memory for that guest. libvirt stores information on the free memory available on each node. Use the virsh freecell --all command to display the free memory on all NUMA nodes " I have tried to look at what freecell returns and heuristically its total across cells looks roughly the same value returned by the 'free' command but without discounting buffers and cache. While free is able to show the value when buffers/cache are taken into account, free is not able to apportion that to NUMA cells. So is it really a valid strategy to choose the NUMA cell for pinning vcpus looking at virsh freecell or ideally we'd need those values adjusted with buffers+cache ? And if so how could one do that ? -------------------------------------------------- regards, Edoardo Comar IBM SWG UK - Hursley JTC ecomar(a)uk.ibm.com Tel +44 (0)1962 81 5576 (external) 245576 (internal) IBM UK Ltd, Hursley, SO21 2JN Unless stated otherwise above: IBM United Kingdom Limited - Registered in England and Wales with number 741598. Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU

1 0

[libvirt-users] virsh list not working with xen 4
by Rogério Vinhal Nunes 21 Sep '12

21 Sep '12

Hi, I'm having some trouble to get libvirt to show the correct power state of my virtual machines. I'm using Ubuntu 10.04 + Xen 4.1.1 + libvirt 0.8.8. virsh list --all only shows turned off machines registered in xend. If I turn them on, they just "disappear", and when I start machines directly from xml, they just doesn't appear at all. Libvirt is correctly connecting to xen as I can use the other commands fine, just the list option doesn't seem to work at all. What can I do to change that? # virsh version Compiled against library: libvir 0.8.8 Using library: libvir 0.8.8 Using API: Xen 3.0.1 Running hypervisor: Xen 4.1.0

2 1