Quick quiz: What stops you from using libvirt?

Libvirt is a great way to manage all sorts of virtualization systems (Xen, KVM, VMWare, OpenVZ, etc). And there are some nice command line and graphical tools built on top of it like virsh and virt-manager. It’s very mature, having been in development for something like 4 years.

But for some reason, not everyone in the world is using libvirt! Why don’t / can’t / won’t you use libvirt? Share your pain …

63 Comments

Filed under Uncategorized

63 responses to “Quick quiz: What stops you from using libvirt?

  1. red

    I use libvirt/virt-manager for private use of KVM, but at work it’s all VMware ESX and I wouldn’t know how to do that…since I don’t have admin privileges on the ESX stuff and don’t know much about it anyway.

    • rich

      We support simple ESX operations, domain start and shutdown, save, restore, that sort of thing.

      Commands are things like:

      virsh -c esx://esxserver/ list --all
      

      You might need to add ?no_verify=1 to the connection URL if you don’t have a certificate infrastructure.

      I have not tried this with virt-manager though …

      • joshuadf

        I didn’t know that. I’ll have to try it out. Does libvirt support ESXi too (WBEM interface)?

      • rich

        joshuadf: Ummmm .. I’m not sure. I know it interacts with ESX using the SOAP API, if that’s the same thing.

      • joshuadf

        Looks like no. If you’re ever interested in adding basic ESXi support, there’s an command line WBEM script at: http://communities.vmware.com/docs/DOC-7170

        On Fedora, just “yum install python-pywbem” and run “python check_esx_wbem.py https://myesxi:5989 root password”.

        ESXi itself is free beer, but the hardware support is somewhat limited (if you have a spare Dell or HP rack server it should work).

  2. mbooth

    I am using libvirt! Virt-manager is particularly awesome.

    I even migrated all my VMware development images to KVM with the move to Fedora 12 because I was tired of the kernel module recompilation dance I had to do with every update.

    Thanks for the top notch (and constantly improving) virtualisation experience…

  3. I use virt-manager / libvirt / kvm for all virtual machine-based Fedora testing I do, and it’s fantastic, so thanks for that.

    There is a situation where I don’t use it. My server setup is a dedicated machine to host two VMs, one mail server and one web server. It runs VMware Server on Mandriva. The reason for this is very simple: I set it up that way quite a while ago – it ran VirtualBox for a while but that didn’t seem to handle two VMs at once very well – and it works extremely well. I don’t have the motivation to convert it to virt-manager / libvirt, especially as they’re not quite as seamless on Mandriva as they are on Fedora, so I’d probably wind up switching the server to running Fedora at the same time.

    I’m perfectly happy with this state of affairs, but you asked for data points so here’s one. If there were some sort of magic one-click system for instantly converting a system running VMware Server on Mandriva into one running virt-manager on Fedora I’d probably switch, but I’m not expecting that to show up🙂

    • rich

      for instantly converting a system running VMware Server on Mandriva into one running virt-manager on Fedora

      By this do you also mean converting from VMWare to KVM as well? We have a tool in Fedora called virt-v2v which do certain sorts of conversion like this, not currently VMWare to KVM unfortunately.

      • No, converting the machine image would probably be the easiest bit. It’s more either getting virt-manager working on Mandriva or switching the host to Fedora, and getting the networking working…

      • rich

        Adam: I think outreach to other distributions is important. We should make sure the tools work well on all the major distros, and not leave that so much up to the distros themselves to get right.

      • Buchan Milne

        virt-manager works just fine on Mandriva. I have a virtual 2-node cluster running Mandriva cooker (pre-2010.1) with GFS under KVM on my Mandriva 2010.0 laptop, using only distribution packages. I previously (Mandriva 2009.0, 2009.1) used a similar setup with Xen (with virt-manager).

        We are missing some tools (libguestfs etc.). Then again, it is a bit difficult to understand which tools to use. libguestfs? With virt-v2v? Does virt-v2v depends on libguestfs? Is Matahari going anywhere, or will this be obsoleted by RHEV? Will the RHEV stuff be open-sourced?

        From a RHEL user perspective, I need snapshot support, but it would be nice to see a feature comparison with the competitors (vSpere, XenServer) showing which are included with RHEL, which require RHEV etc.

      • rich

        We are missing some tools (libguestfs etc.). Then again, it is a bit difficult to understand which tools to use. libguestfs? With virt-v2v? Does virt-v2v depends on libguestfs? Is Matahari going anywhere, or will this be obsoleted by RHEV? Will the RHEV stuff be open-sourced?

        Yes, virt-v2v depends on libguestfs.

        Matahari (the guest agent) is being actively developed.

        RHEV will be (or already is?) open sourced. I’ve not been paying much attention to the Windows side, but Red Hat is committed to this.

  4. pbor

    (I admit I have not yet tried the latest and greatest virt-manager)

    One thing I really missed last time I tried and which keeps me to vmware is good cut and paste support between the host and the guest. With vmware I can cut and paste text, files and even drag and drop them

    • John Florian

      I agree this feature, however implemented would be awesome. Once I have the guest setup, I use ssh and then cut-n-paste from host to guest is trivial. I want cut-n-paste before I get that far. For example, it would be wonderful for doing the initial network configuration.

      • rich

        I think seamless cut and paste, video streaming, USB, microphones etc is what SPICE will solve for us, once it goes upstream.

  5. Nik Lam

    We’re using it on RHEL 5 to do our virtualisation. One thing that I think would help get more early-adopters would be more documentation, especially howto/tutorial type stuff.

    Here’s an example of a couple of gaps in the doco that I’ve hit:

    (1) When using Xen, there was doco out there for how to dedicate an ethernet interface for live migration purposes. However now we’ve switched to KVM where libvirt handles the live RAM transfer, I’m struggling to acheive this.

    (2) I had trouble when I wanted to set up some firewall rules to protect the host. I was allowing tcp connections to libvirtd’s TLS port 16514 but live migrations were failing. It was relatively difficult to find out that the RAM migration happens over a new connection on ports ranging between 49152:49216. Once I opened up this range, the migrations worked.

    I guess an article on how live migration is handled by libvirt, including the interactions between the hosts involved etc would have been useful. Also it looks like maybe the virsh “migrateuri” option might be a clue, but the documentation of it is pretty minimal (“migration URI, usually can be omitted”).

    Lastly, I think the lack of integration between libvirt, encryption/TLS and VNC is quite challenging if you want to secure your network traffic. There are too many components and steps required to allow a quick and easy setup. At least it’s mostly documented though🙂

    Generally though, I’m pretty happy and I’ve been impressed by the progress I’ve been seeing in each successive Fedora release. Thanks for the blog posts and for asking for feedback too!

    • rich

      Very well-reasoned points here Nik. More / better-organized documentation is obviously needed. This was the original idea behind the libvirt wiki, but the information there now is very incomplete. There is nothing at all about migration, for example, never mind the detailed issues you raise about port numbers and firewalls.

  6. I think the proper question is why is someone not using libvirt not whether they are using it indirectly.

    I used it but I didn’t use libvirt remote, for instance. I didn’t use the storage features. I didn’t use the networking features. I think the library is growing well into the “manage entire operating system” space and possibly beyond an ideal charter of small tools working together, where some of those developments could be better used to manage even physical OS’s, and should be so abstracted and not kept in the virt realm.

    I didn’t like that it that client libraries to simplify interaction with libvirt frequently lag behind the XML. Further shiny new XML features are not available as backports on older versions of the OS (for obvious reasons), so when coding towards enterprisey things it is difficult to become excited about such newness from a tools/business perspective, knowing how long it is until they reach RHEL.

    The problem of not having strong/stable client libraries means a libvirt developer has to more closely engaged with internals of the libvirt community, which can be a fair amount to keep up with. It moves too quickly, IMHO.

    Obviously this is how innovation happens, and the goal is to have that innovation when EL 6 is available, though this may indicative of why it is slow to be adopted.

    For many simple virt tasks, just using the command line tools goes almost all the way, so I can see a lot of folks being content to use / shell out to those rather than track libvirt closely.

    In all the infrastructure is of course great, but if you’re asking about why there aren’t more people using the library, it may because it’s too easy to /not/ use it.

    • rich

      The problem of not having strong/stable client libraries means a libvirt developer has to more closely engaged with internals of the libvirt community, which can be a fair amount to keep up with.

      And I think the difficulty that a lot of the code is “string parsing in C” which makes my head spin too.

      For many simple virt tasks, just using the command line tools goes almost all the way, so I can see a lot of folks being content to use / shell out to those rather than track libvirt closely.

      It’s a shame that people don’t use the language bindings more often. libvirt is much easier to use from Python/Perl, than from shell or C.

      • Indeed, though I think what I’m getting at is the threat of python-virtinst (and the like) going away.

        Sys admins often write rather clever software, but are usually allergic to XML.

        I’d like to see that library become fully featured, and to possibly have a Ruby analog (and Perl too, maybe).

        That’s a lot to maintain, but it would help.

    • Daniel Berrange

      > I think the library is growing well into the “manage
      > entire operating system” space and possibly beyond an
      > ideal charter of small tools working together, where
      > some of those developments could be better used to
      > manage even physical OS’s, and should be so
      > abstracted and not kept in the virt realm.

      The reason the API has grown beyond just management of VMs, and added storage + networking is because that is neccessary in order to be able to manage hypervisors such as VMWare / Hyper-V / VirtualBox. While with Xen & KVM you had Linux underneath, so could easily use standard Linux mgmt tools/APIs for storage & networking like HAL/DeviceKit/etc, you cannot do the same with other hypervisors, whose mgmt OS is a black box you can only access via their APIs. Thus the libvirt API has to grow to allow for full support for storage/network mgmt in VMWare / Hyper-V / etc in a standard API.

  7. fkooman

    I’m using VirtualBox as opposed to KVM/QEMU/libvirt because I’m using desktop virtualization which is not as smooth in KVM/QEMU/libvirt because of the lack of dynamically resizing display support.

    Maybe it will get better with SPICE?

  8. Stijn Hoop

    Me and my colleague switched to KVM from VMware Workstation just now as well, on Fedora 12. We don’t have that much use for most of the libvirt features as we usually use local virtual machines anyway. I did try to use the new netcf (not sure about the libvirt integration there) to set up a bridge but it was still not a smooth experience mostly due to NetworkManager not understanding bridges at all.

    We do have a large ESX system as well but I haven’t yet tried to manage stuff on it with libvirt. Mostly a combination of lack of time on my part and lack of very accessible documentation I think.

    Hope this helps as another datapoint. I don’t mean to sound negative, in fact I really like the virtualization features in F12. Any transition will take some time to get used to all the new stuff though, for me at least🙂

  9. Leif Gruenwoldt

    I’ve been a user of vmware server and now ESX for several years now. As you probably already know VMWare has terrible ESX user/client support on linux right now. There is no native linux tool for connecting or managing VM’s for ESX (there was for vmware-server 1.x). The user has two optoins:

    a) use the clunky ESX firefox web plugin (that only works in specific versions of firefox)

    b) hack apart the VMWare Server 2.x firefox plugin and pulling out a GTK client to run standalone. This actually works relatively well but can only connect to a VM, not manage it. ( http://leif81.blogspot.com/2008/12/vmware-remote-console-for-esx-on-linux.html )

    Neither of these options allow management of the VM’s (most importantly managing snapshots). My solution for this problem after discussing this on #vmware was to create a Windows VM, install ESX VC Client on it, and connect to that with my hacked up GTK client.

    So long story short, the reason I don’t use libvirt yet is it doesn’t make my above problem any better. I’m stuck with ESX for the time being and I just want to connect to my ESX VM’s and manage them (from a GTK client I can install from yum!)

  10. I haven’t used libvirt until recently, mostly because it wanted to force me to use NATed network (with iptables and dnsmasq) while I manage already my own network support.

    But there is more to that I’m afraid, as a packager for Gentoo I feel that too much is only designed to work for RedHat or Fedora and the way they (you) use it. See for instance my rant about the XML files not being easy to detect, and thus edit with some support, as they lack any type of namespace, declaration, specific name of tags and so on (it’s on libvir-list if you don’t want to dig it up on my blog).

    And while I’m trying to do my part by submitting patches, both for fix and improvements, it doesn’t help that the one (or maybe first) time a patch I suggested “break” a (non-default) configuration I have to defend myself against the idea that I “invalidated the FAQ”.

    • rich

      I believe that you’re not forced to use the default NATed network. You can delete the default network (virsh net-undefine default). However you are right that this is not exactly an intuitive thing, and furthermore you have to replace it with something else which can be even harder to get right.

      For the rest — see my comment above about being better at helping other distributions to package our code.

      And thanks for doing the Right Thing by engaging and contributing to libvirt.

      • Well, I tried it again with 0.7.4 (*cough* patched — the original 0.7.4 release was unusable as it crashed almost every time) and it at least left me the option to use my own (already configured) bridge, so I’m fine and dandy for that right now.

        I know that other (Gentoo) users have had the same reserve against libvirt: we’re used to have mostly-manual but very flexible configuration for networking in Gentoo, so the all-or-nothing of the older versions seemed very out of place for many of us.

        In my TODO list for what concerns integration right now is to get init scripts to not restart the running domains when restarting libvirtd, and for handling the domains themselves. But as you’ve seen, I don’t refrain from getting code out.

        It _definitely_ has improved in the past year I’d say.

  11. David Cartwright

    I am having a big problem with libvirt + bridging + KVM when the VMs are set to autostart.

    Bugzilla report is here:
    https://bugzilla.redhat.com/show_bug.cgi?id=533087

    Basically, if IPv6 is enabled, autostarting VMs on both Fedora 11 and Fedora 12 systems (64-bit) causes hard lock of system. Since the systems work fine if started manually, I suspect the problem is related to libvirt and not the kernel.

    You said “share your pain” … and that’s mine!

  12. Also +10000 to Network Manager understanding bridges ASAP. Outside connectivity into guests is a must, and these are a /pain/ to set up manually.

    I really liked the days when xenbr0 was available as a real bridge, auto-created by Xen. virbr0 not being a bridge is a frequent stumbling block by many end users.

    • rich

      I didn’t like debugging all the problems that Xen created by munging the network configuration using large, non-integrated boot-time scripts.

    • Daniel Berrange

      libvirt now has the ability to configure all types of networking (bridges, bonding, VLANs, ipv6, etc). This is targetted to be integrated in virt-manager, which will help this problem significantly. In addition NetworkManager will really soon have support for bridging, though bonding + vlans will be further off on its roadmap.

  13. …because the UI sucks!
    🙂

    Not meant as offense, I really like and appreciate the improvements, but it’s still lacking the focus to detail. Have you ever tried to have conducted an user interaction testing with a real human? VirtManager is probably most weirdly behaving application that I’ve ever used, I’ve always thought it’s just unfinished, but now that you ask why don’t people use it yet I’m a bit puzzled…

    To be fair, my answer is technically not answer to why I “don’t use it yet,” since I, and all my colleagues, use it on a daily basis and it served use very well so far.

    Just a few random details that me and my colleagues find irritiating (warning, lot of personal preferences below):

    1.) Send key is not able to send arbitrary keys
    2.) No cold reset of guest vm
    3.) VNC/serial console/details are switched via radio buttons in menu? tabs would be more natural
    4.) Multiple windows for manager and VMs? Vinagre uses one, and its gui is nicer.
    5.) Saving of snapshots is awkward. See what VirtualBox does with the tree graph of snapshots.
    6.) “Add Hardware” button in a strange place. Many colleagues had hard times finding it🙂 How about making the hardware list right-button clickable?
    7.) Add hardware wizard feels so windows 95’s dialogs with far too many steps. When I select that I’m adding disk, why do I have to confirm it with “Forward”, this could all be dealt in one dialog.

    • rich

      Thanks for the constructive feedback and good suggestions. I’ve passed it along to Cole.

    • Hi Lubomir,

      Thanks for the virt-manager feedback.

      1) A flexible send key option has been on the todo list for a long time, but it has always been bumped by higher priority work.

      2) We could implement this in virt-manager, but this is really a libvirt issue: there is no reboot implementation for the qemu driver.

      3) VNC vs. HW Details was done with tabs until F12, but this was deliberately dropped. Since most users will presumably be spending most of their time at the VNC window, a tab bar was considered a bit overkill WRT screen real estate. There is a toolbar in the details window with buttons for this (maybe you have that hidden).

      Serial console handling is a bit wonky, I agree. However, it’s a pretty niche requirement: sticking the choice in either a tab bar or toolbar is overkill since at least 95% (my guess) of kvm virt users won’t even have a serial console configured properly. We could possibly make the toolbar configurable with an optional serial console button. Any other ideas are welcome.

      4) Hmm. It is suboptimal in some ways no doubt, but a comparison to vinagre isn’t very fair, we obviously have to expose much more functionality then just VNC. I think there would be value in a single window that allows easily switching between multiple VMs, just not sure how the best way to go about it.

      5) Libvirt doesn’t really have proper snapshot support ATM. You are probably talking about save/restore behavior, which is more like ‘suspend to ram’. The UI for this is indeed crappy right now, but properly improving this is largely blocked on libvirt implementation.

      6) Never heard that before. I think adding a right click ‘add hardware’ option is a good idea.

      7) No argument there🙂 The precedent for this was set years ago and it has grown from there into quite the beast. We’ve been doing UI redesigns of virt-manager pieces incrementally, just haven’t got to the Add Hardware wizard yet.

      • Daniel Berrange

        > 4) Hmm. It is suboptimal in some ways no doubt,
        > but a comparison to vinagre isn’t very fair, we
        > obviously have to expose much more functionality
        > then just VNC. I think there would be value in a
        > single window that allows easily switching
        > between multiple VMs, just not sure how the best
        > way to go about it.

        With firefox everything appears in tabs, but you have the option to detach a tab creating a new top level window out of it. This paradigm might work nicely in virt-manager too.

  14. I think the main reason is we’re using Xen and we’re used to xm CLI.

  15. Anonymous

    We use Xen and have begun managing some of our Xen domains using libvirt. Not quite as robust as ‘xm’ yet but very usable.

  16. I don’t use libvirt, because it is too hard to set it up to run it as non root. E.g. I just installed it to test it again. I know there is some URI that is supposed to allow using virt-manager with a non root user libvirtd instance, but if I open “add connection”, there is no option to do so easily.

    I do not like to run libvirtd, because it seems to mess around a lot in my system. I am not sure what the current status is, but it used to create it’s own bridge by default and I read about it changing iptables rules. This might be useful for people who do not care about their system, but I would like to have control about what happens on my system.

    Also the last time I checked libvirtd used a network connection for the vnc console instead of a unix domain socket, which is imho the right way to do for local connections.

    What I would like to have is some way to run libvirtd easily as a non root user, but still have it suspend virtual machines on power down of the host machine.

  17. Chris

    From the libvirt/kvm perspective:

    1) No support for proper snapshots. This is a killer for real world deployment – this company has a requirement that testers need to be able to snapshot and restore from a GUI, and virt-manger can’t do it.
    2) Reboot VM does not work. How hard can it be to implement?
    3) virt-manager can not create a VM with physical (bridge) interface when connected remotely. Another show stopper for real-world deployment – our testers need to be able to create VMs attached to the network. It is possible to ssh in and run virt-manager over X11, but then what is the point of the remote connection facility?
    4) This is a sucky complaint, but there are Windows users here who want/say they need a Windows virt-manager or similar… they do not like the idea of remote virt-manager over X11.

  18. Shane Falco

    I use libvirt quite a bit. In our large xen environment I mostly develop our processes now using libvirt with the future idea that we may perhaps move to KVM and not have to rewrite everything.

    But it’s a love-hate relationship, mostly hate because the standard Xen tools and interfaces work better and more simply…like they were written by an actual user of the software. The libvirt “backend” (ie, not the commands like virt-top, virt-manager, guestfish, etc.), otoh, feels like it’s designed by committee. I despise XML for configuration. Every unix guy I know despises XML for configuration. I will never understand why someone who’s clearly an amazing developer (I love this blog) would use garbage like XML for something that clearly needs to be easily scripted.

    • rich

      To be fair, the use of XML in libvirt is nothing to do with me, and not my favorite decision either.

      • Shane Falco

        That would explain it, then! Perhaps my “designed by committee” feeling wasn’t too far off.

        By the way, I know you know, but I’m going to say it anyway. Your virt tools rock. It’s people like you who keep Fedora/RHEL ahead of everyone else.

      • Anonymous

        it’s becasuse someone in the development team is the maintainer and author of libxml🙂

  19. Al Tobey

    I have about 60 servers running libvirt 0.7.6 (just upgraded) and KVM.

    The number one annoying problem for me is that I cannot import/abstract bridges that were created by the OS configuration mechanism. That means I still have to configure domains to directly hit the bridge, which reduces the power of libvirt IMO.

    Ideally I’d like to be able to use network=”foo” and underneath have libvirt map that to br999 or whatever (I map 1:1 bridge->802.1q interface).

    It would also be nice to be able to do similar things with storage. I can’t say a lot about that yet, but I have 50/60 servers using LVM via libvirt and it’s working OK so far. In this case, the “manage everything” style of the API is quite handy since it makes managing all the boxes from my Django app quite a bit easier.

  20. Erik de Castro Lopo

    The main problem for me is that libvirt doesn’t support VDE2 networks for qemu/kvm machines.

  21. What stops me using virt-manager is that it has a bug working with XEN dom-Us, where the built-in vnc display would show garbled screen of the domU. It happens when I open-close-open the DomU from virt-manager. I ve found 2 ways to workaround: 1) Reboot the VM (nasty, we`re working with servers);
    2) fire a vncviewer from tightvnc to connect to the vm, it would fail with “..Unknown rect encoding..”, then I run it again (it would connect o.k.), close it and open the domU again in the virt-manager. The display is o.k. then.
    libvirt is nice, but it`s the next layer above the hypervisor`s own tools (xend/xm in my case) so more layers=more potential problems/bugs/failures. In centos5.1 it was not very stable so i made my own OCF sripts for HA which use xm commands to manage the guests (domUs in xen).

  22. I run command line only servers without X-windows installed. I need the QEMU-KVM graphics equivalent to -curses.

    I really, really, want to use Libvirt, in no small part for clean startups and shutdowns, but need this functionality. Am I missing it? VNC doesn’t help me since that too appears to need X.

    • rich

      virsh console? I use libvirt all the time on headless machines. It shouldn’t require X, although obviously tools like virt-viewer need X by their very nature.

  23. Trying virsh console, but all I get is “Connected to domain Test” “Escape character is ^]” – with no console output.

    Perhaps I should back up. Without libvirt, I would start the KVM instance like this to have a console (handy when doing initial installs before the network is configured):
    kvm -net nic,model=virtio -net tap,ifname=tap0,script=no -drive file=test.img,if=virtio,boot=on -curses -no-reboot -m 1g -smp 2
    That would in turn create a curses based window and I’d see and be able to interact with everything from the Grub prompt onwards. Later, when stable, I’d skip the “-curses -no-reboot” and use “-daemonize” instead, and connect via ssh.

    I’m found virt-install, and am guessing this command should get something similar defined within Libvirt:
    virt-install –name “Test” –ram 1024 –vcpus=2 –cpu host –description “Test VM” –import –file test.img –os-type=linux –os-variant=virtio26 –network bridge=br0,model=virtio –graphics=none –autostart –prompt
    I can then escape out like with virsh. No boot messages are generated.

    “virsh list” shows the domain “Test” active, and “virsh edit Test” shows me the xml file. That felt reasonable. Top shows an instance running, but its not consuming much.

    I’ve tried this with a copy of my boinc vm, which is fully daemonized and sucks 6 cpus dry when autostarted via the kvm command in my local.d/local.start file (Gentoo Linux distribution btw). The started “Boinc” domain show as running under top, with a virtual memory size as set, but with only 29m resident and 0-1% CPU. Sample output:

    virsh # list
    Id Name State
    ———————————-
    4 Boinc running

    virsh # console Boinc
    Connected to domain Boinc
    Escape character is ^]

    virsh # destroy Boinc
    Domain Boinc destroyed

    I’m guessing I’m missing something VERY basic, since my libvirt knowledge is at the newbie level. Any help would be greatly appreciated.

  24. ps. I’m guessing the virt-install network line should be “-network bridge=tap0” rather than the actual br0 device, but wouldn’t expect that to affect the console issue.

  25. Thanks. Trying and seeing:
    virt-install –name “Boinc” –ram 4096 –vcpus=6 –cpu host –description “BOINC CPU sucker” –import –file boinc_libvirt.img –os-type=linux –os-variant=virtio26 –network bridge=br0,model=virtio –graphics=none –autostart –prompt -x “console=ttyS0,155200”
    This will overwrite the existing path ‘/kvms/boinc_libvirt.img’
    Do you really want to use this disk (yes or no)
    yes
    ERROR –extra-args only work if specified with –location.

    Tried various options using –location instead of –import, but Gentoo doesn’t offer such an installable distribution that I’m aware of. I tried specifying the LiveDVD.iso, a minimal install iso, and one of my base image files – but that was just guessing on my part.

    Do sincerely appreciate the attempts at helping me through this.

  26. I’m comfortable editing a previously created domain xml file to add whatever
    -x “console=ttyS0,115200″ would have added if you could enlighten me as to what that would have been. Might be the easiest way!

  27. Thanks. I’ll go check the previous post!

  28. OK – thanks everyone. Making progress. But first… should I take this conversation to the mailing list?

    Progress: I mentioned earlier that virsh showed my KVM being launched, but “top” didn’t show any significant resource usage. After trying several terminal emulators and even trying a telnet port for access, I decided to do the obvious: go to /var/log/libvirt/qemu and see what startup command was being issues. The very first thing I noticed was qemu-kvm being passed the “-S” flag. That explains why nothing is happening! A quick test by hacking the startup by changing “nographics” to “curses” confirmed no activity upon boot as expected. Take out the “-S” and we are running. I have other problems, but need to get libvirt to stop using “-S” first.

  29. tomás zerolo

    It’s the XML.

    I’m starting to play with KVM, privately to gain experience. At the moment I try to avoid libvirt like a plague, because as a matter of policy I try to avoid software with XML configuration files.

    I know it’s popular at RH😦

    • rich

      In most common circumstances it should not be necessary to write XML directly.

    • Craig

      Seconded. I’d much rather do things manually than deal with XML. It really doesn’t matter who has to write it. The pointless complexity tax has to be paid by someone, even if not me. It all takes it’s toll. It’s disingenuous to tell programmers something to the effect of “Oh don’t worry about it sir, that’s all taken care of by us. We’ll keep this piece of crap running smoothly”.

  30. jcoffland

    libvirt STILL does not support vde. Also the documentation is very fragmented.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s