[fedora-virt] KVM high available

Dor Laor dlaor at redhat.com
Tue Jan 18 16:29:30 UTC 2011


On 01/18/2011 02:06 PM, iarly selbir wrote:
> I configured a Gluster volume to work as my backend storage instead gfs
> but I can't finish the setup of a vm, after click em
> finish(virt-manager) it show me this error:
>
> Unable to complete install 'libvirt.libvirtError internal error unable
> to start guest: qemu: could not open disk image
> /var/lib/libvirt/images/test.img
>
> Traceback (most recent call last):
>    File "/usr/share/virt-manager/virtManager/create.py", line 724, in
> do_install
>      dom = guest.start_install(False, meter = meter)
>    File "/usr/lib/python2.4/site-packages/virtinst/Guest.py", line 541,
> in start_install
>      return self._do_install(consolecb, meter, removeOld, wait)
>    File "/usr/lib/python2.4/site-packages/virtinst/Guest.py", line 633,
> in _do_install
>      self.domain = self.conn.createLinux(install_xml, 0)
>    File "/usr/lib64/python2.4/site-packages/libvirt.py", line 974, in
> createLinux
>      if ret is None:raise libvirtError('virDomainCreateLinux() failed',
> conn=self)
> libvirtError: internal error unable to start guest: qemu: could not open
> disk image /var/lib/libvirt/images/test.img
>
> # /var/log/messages
>
> Jan 18 05:36:05 kvmsrv001 libvirtd: 05:36:05.227: error : internal error
> Timed out while reading console log output
> Jan 18 05:36:05 kvmsrv001 libvirtd: 05:36:05.227: error : internal error
> unable to start guest: qemu: could not open disk image
> /var/lib/libvirt/images/test.img
>
>
> The file test.img is created but the domain is not, If I umount
> /var/lib/libvirt/images (mount point to my gluster volume) all works fine.
>
> Does anyone experienced with this?

Seems like there is some sort of an access/permission issue.
Since your disk should be shared, make sure that it's there in the 
libvirt xml and the file is writable.

>
>
> - -
> iarlyy selbir
>
> :wq!
>
>
>
> On Tue, Jan 18, 2011 at 3:52 AM, wariola at gmail.com
> <mailto:wariola at gmail.com> <wariola at gmail.com
> <mailto:wariola at gmail.com>> wrote:
>
>     I think u can use cluster suite and fence towards Libvirt as a
>     fencing device.
>
>     (Tho I just know the theory, never tried it myself)
>
>     -wariola-
>
>     On Mon, Jan 17, 2011 at 4:17 PM, Dor Laor <dlaor at redhat.com
>     <mailto:dlaor at redhat.com>> wrote:
>
>         On 01/14/2011 08:34 PM, iarly selbir wrote:
>          > Sorry, I forgot to mentions it, Yes I have this configuration two
>          > failover domain, first with host-1 on top, and other with
>         host-2 on top.
>          >
>          > My question is, how you guys are configuring your guests
>         resources to
>          > failover from on hosts to another, remembering that I have
>         same machines
>          > on two kvm hosts, i.e. kvm001 has guest001 on, and kvmsrv002 has
>          > guest001 ( It must be powered on just in fail of guest0001 on
>         kvm001)
>
>         Using hearbeat/light cluster mgmt ala Linux HA package might be
>         a good
>         option for you.
>
>          >
>          > I hope being clear enough.
>          >
>          > Thank you so much.
>          >
>          > - -
>          > iarlyy selbir
>          >
>          > :wq!
>          >
>          >
>          >
>          > On Fri, Jan 14, 2011 at 3:18 PM, Thomas Sjolshagen
>          > <thomas at sjolshagen.net <mailto:thomas at sjolshagen.net>
>         <mailto:thomas at sjolshagen.net <mailto:thomas at sjolshagen.net>>>
>         wrote:
>          >
>          >     On Fri, 14 Jan 2011 14:19:52 -0300, iarly selbir
>         <iarlyy at gmail.com <mailto:iarlyy at gmail.com>
>          > <mailto:iarlyy at gmail.com <mailto:iarlyy at gmail.com>>> wrote:
>          >
>          >>     Hi there,
>          >>     Hi I'm joining today an would like to share my knowledge
>         with
>          >>     virtualization and get more = )
>          >>     when KVM-HOST-001 fail, the KVM-HOST-002 take over all
>         machines
>          >>     from other host, I'm sharing a storage volume between
>         two nodes
>          >>     (gfs2), so all hosts can see the guest images, but how to
>          >>     configure the clusters resources to migrate the guests?
>         this is my
>          >>     question and any suggestions will be appreciated.
>          >     Assuming you're using libvirt to manage the VM's
>         (guests), I'd
>          >     configure them as <vm> resources in rgmanager and make
>         them members
>          >     of a failover group with the highest (shows up as the lowest
>          >     priority number in the example) priority to the
>         KVM-HOST-* you want
>          >     the guest to start on (if it's available).
>          >     A couple of (example) <vm> resource I have configured in
>         my 2-node
>          >     GFS2 based KVM cluster (some of the info in the vm
>         resource tag is
>          >     actually not necessary, but I was both experimenting and
>         playing it
>          >     safe when I set this up)
>          > <rm>
>          > <failoverdomains>
>          > <failoverdomain name="prefer-virt0" restricted="0" ordered="1">
>          > <failoverdomainnode name="virt0-backup" priority="10" />
>          > <failoverdomainnode name="virt1-backup" priority="20" />
>          > </failoverdomain>
>          > <failoverdomain name="prefer-virt1" restricted="0" ordered="1">
>          > <failoverdomainnode name="virt1-backup" priority="10" />
>          > <failoverdomainnode name="virt0-backup" priority="20" />
>          > </failoverdomain>
>          > </failoverdomains>
>          > <!-- VM resources -->
>          > <vm name="imap1" autostart="1" recovery="restart" migrate="live"
>          >     domain="prefer-virt0" use_virsh="1" hypervisor="qemu" />
>          > <vm name="imap2" autostart="1" recovery="restart" migrate="live"
>          >     domain="prefer-virt1" use_virsh="1" hypervisor="qemu" />
>          > </rm>
>          >     Hope this helps to illustrate.
>          >     // Thomas
>          >
>          >
>          >
>          >
>          > _______________________________________________
>          > virt mailing list
>          > virt at lists.fedoraproject.org
>         <mailto:virt at lists.fedoraproject.org>
>          > https://admin.fedoraproject.org/mailman/listinfo/virt
>
>         _______________________________________________
>         virt mailing list
>         virt at lists.fedoraproject.org <mailto:virt at lists.fedoraproject.org>
>         https://admin.fedoraproject.org/mailman/listinfo/virt
>
>
>
>
>     --
>     .: war|ola :.
>     Use Fedora Linux for better computing experience
>     http://fedoraproject.org
>
>



More information about the virt mailing list