Installing a VM in oVirt Appliance Host: Dell Optiplex 745, VT-capable, x86_64 Nodes: four cells of ES7000/one, each with 8x at 3.4GHz and 32GB memory Downloaded and installed 0.96 release on top of Fedora10 as indicated in the documentation. Successfully PXE booted 4 ES7000 nodes, and all show up in the oVirt dashboard. Defined a storage pool using the /ovirtnfs share from the appliance at 192.168.50.2 Created a Virtual Machine using one of the NFS disks, 4cpu and 1024mb However, I'm unable to install into that VM. If I launch the QEMU viewer as the VM launches, I can select the Fedora option from the PXE installer... but the fedora install always reports an error with the network interface. If I select retry, the install goes a little farther, downloading the install.img and running the installer. However the system hangs for a long time at "Retrieving installation information for f10-x86_64-updates..." and then fails at "Unable to read package metadata..." Any ideas? Also, I saw someone mentioned installation from an ISO - is this supported? Is it possible to use existing VM images? I might be able to specify "boot from HD" and then dd an existing image - but i was looking for a more official method :-) brian carb unisys corporation - malvern, pa brian.carb at unisys.com
Carb, Brian A wrote:> Installing a VM in oVirt > > Appliance Host: Dell Optiplex 745, VT-capable, x86_64 Nodes: four cells > of ES7000/one, each with 8x at 3.4GHz and 32GB memory > > Downloaded and installed 0.96 release on top of Fedora10 as indicated > in the documentation. Successfully PXE booted 4 ES7000 nodes, and all > show up in the oVirt dashboard. Defined a storage pool using the > /ovirtnfs share from the appliance at 192.168.50.2 Created a Virtual > Machine using one of the NFS disks, 4cpu and 1024mb > > However, I'm unable to install into that VM. If I launch the QEMU > viewer as the VM launches, I can select the Fedora option from the PXE > installer... but the fedora install always reports an error with the > network interface. If I select retry, the install goes a little > farther, downloading the install.img and running the installer. However > the system hangs for a long time at "Retrieving installation > information for f10-x86_64-updates..." and then fails at "Unable to > read package metadata..."The initial network error you get that is solved with a retry I have seen every time I try to provision Fedora from a kickstart in oVirt. Hitting retry has always fixed it for me. The problem is probably a transient network issue, like the network bridge not being fully up and forwarding traffic by the time the guest gets to this phase of the install. If you have bugzilla access and can file a bug at bugzilla.redhat.com under Other->Virtualization Tools->ovirt-node we'll hunt this down and try to fix it. After the retry though provisioning generally works for me... It could be a few things... The way we have the appliance set up, we provision from cobbler but the cobbler repos are just pointers to upstream Fedora mirrors. So first thing to check would be to make sure that from a Node you can access (ping) the internet. If not, that's probably the issue. If you can, it could be a mirror problem since sometimes mirrors are unreliable.> Any ideas? > > Also, I saw someone mentioned installation from an ISO - is this > supported? Is it possible to use existing VM images? I might be able to > specify "boot from HD" and then dd an existing image - but i was > looking for a more official method :-)We do support provisioning from ISO files. There should be something on the wiki about this, if not Darryl Pierce is the developer who set that up. The short of it is that you add an ISO image to the /cobblernfs directory on the Appliance. Once the ISO is there you need to use the cobbler image add command to add the ISO image to cobbler This will make the image ISO show up in the oVirt UI under the provisioning drop down. Try that out and let me know if it works. For disk image provisioning, we have plans to support that as well. Just not implemented yet... Perry
Carb, Brian A wrote:> I added a slax iso and then did the image add in cobbler. Then, when I create a VM, the slax cobbler ISO entry showed up in the interface drop-down as you said it would, and so I selected it. But the VM refuses to start - taskomatic.log shows a libvir error: > > libvir: Storage error : XML description for missing storage pool source host name is not well formed or invalid > > i have /ovirtnfs added as a storage pool. I tried adding /cobblernfs but this didn't seem to fix anything. > > any ideas?Can you give me the exact syntax for cobbler that you used to add the image? The syntax for the nfs:// url part needs to follow some strict conventions in order to work properly, but upstream cobbler doesn't do any validation of that string so it won't tell you if you did it incorrectly. This thread on ovirt-devel is useful and will probably help fix your issue: https://www.redhat.com/archives/ovirt-devel/2008-December/msg00234.html Perry> > brian carb > unisys corporation - malvern, pa > brian.carb at unisys.com > > -----Original Message----- > From: Perry Myers [mailto:pmyers at redhat.com] > Sent: Friday, January 30, 2009 2:04 PM > To: Carb, Brian A > Subject: Re: [Ovirt-devel] Installing a VM in oVirt 0.96 > > Carb, Brian A wrote: >> Thanks perry. Adding the image to cobbler makes it available - now i'll try to install into it. >> >> I noticed also that after sitting overnight, 2 of my 4 nodes show as unavailable(enabled). I can access them via their consoles though, so the machines are up. any ideas ? >> > > That could be issues with qpid which is the messaging transport we use. I would email Ian Main on ovirt-devel (his nick on irc is Slower) and outline the issue there as he'll have a much better idea of what is going on... > > Thanks, > > Perry > > >> brian carb >> unisys corporation - malvern, pa >> brian.carb at unisys.com >> >> -----Original Message----- >> From: Perry Myers [mailto:pmyers at redhat.com] >> Sent: Thursday, January 29, 2009 8:38 PM >> To: Carb, Brian A >> Cc: ovirt-devel at redhat.com >> Subject: Re: [Ovirt-devel] Installing a VM in oVirt 0.96 >> >> Carb, Brian A wrote: >>> Installing a VM in oVirt >>> >>> Appliance Host: Dell Optiplex 745, VT-capable, x86_64 Nodes: four >>> cells of ES7000/one, each with 8x at 3.4GHz and 32GB memory >>> >>> Downloaded and installed 0.96 release on top of Fedora10 as indicated >>> in the documentation. Successfully PXE booted 4 ES7000 nodes, and all >>> show up in the oVirt dashboard. Defined a storage pool using the >>> /ovirtnfs share from the appliance at 192.168.50.2 Created a Virtual >>> Machine using one of the NFS disks, 4cpu and 1024mb >>> >>> However, I'm unable to install into that VM. If I launch the QEMU >>> viewer as the VM launches, I can select the Fedora option from the >>> PXE installer... but the fedora install always reports an error with >>> the network interface. If I select retry, the install goes a little >>> farther, downloading the install.img and running the installer. >>> However the system hangs for a long time at "Retrieving installation >>> information for f10-x86_64-updates..." and then fails at "Unable to >>> read package metadata..." >> The initial network error you get that is solved with a retry I have seen every time I try to provision Fedora from a kickstart in oVirt. Hitting retry has always fixed it for me. The problem is probably a transient network issue, like the network bridge not being fully up and forwarding traffic by the time the guest gets to this phase of the install. >> >> If you have bugzilla access and can file a bug at bugzilla.redhat.com under Other->Virtualization Tools->ovirt-node we'll hunt this down and try to fix it. >> >> After the retry though provisioning generally works for me... It could be a few things... The way we have the appliance set up, we provision from cobbler but the cobbler repos are just pointers to upstream Fedora mirrors. So first thing to check would be to make sure that from a Node you can access (ping) the internet. If not, that's probably the issue. >> If you can, it could be a mirror problem since sometimes mirrors are unreliable. >> >>> Any ideas? >>> >>> Also, I saw someone mentioned installation from an ISO - is this >>> supported? Is it possible to use existing VM images? I might be able >>> to specify "boot from HD" and then dd an existing image - but i was >>> looking for a more official method :-) >> We do support provisioning from ISO files. There should be something >> on the wiki about this, if not Darryl Pierce is the developer who set >> that up. The short of it is that you add an ISO image to the >> /cobblernfs directory on the Appliance. Once the ISO is there you >> need to use the cobbler image add command to add the ISO image to >> cobbler >> >> This will make the image ISO show up in the oVirt UI under the provisioning drop down. >> >> Try that out and let me know if it works. >> >> For disk image provisioning, we have plans to support that as well. Just not implemented yet... >> >> Perry > > > -- > |=- Red Hat, Engineering, Emerging Technologies, Boston -=| > |=- Email: pmyers at redhat.com -=| > |=- Office: +1 412 474 3552 Mobile: +1 703 362 9622 -=| > |=- GnuPG: E65E4F3D 88F9 F1C9 C2F3 1303 01FE 817C C5D2 8B91 E65E 4F3D > |-=|-- |=- Red Hat, Engineering, Emerging Technologies, Boston -=| |=- Email: pmyers at redhat.com -=| |=- Office: +1 412 474 3552 Mobile: +1 703 362 9622 -=| |=- GnuPG: E65E4F3D 88F9 F1C9 C2F3 1303 01FE 817C C5D2 8B91 E65E 4F3D -=|
Carb, Brian A wrote:>> Can you give me the exact syntax for cobbler that you used to add the image? >> The syntax for the nfs:// url part needs to follow some strict conventions in >> order to work properly, but upstream cobbler doesn't do any validation >> of that string so it won't tell you if you did it incorrectly. > >> This thread on ovirt-devel is useful and will probably help fix your issue: >> https://www.redhat.com/archives/ovirt-devel/2008-December/msg00234.html > > I added an iso image to cobbler with the command > cobbler image add --name=slax-6.0.6 --file=/cobblernfs/slax-6.0.6.iso > > but once I read the thread I issued 2 new commands: > cobbler image remove --name=slax-6.0.6 > cobbler image add --name=slax-6.0.6 --file=192.168.50.2:/cobblernfs/slax-6.0.6.iso --image-type=iso > > but now cobbler daemon cannot restart and complains at every command: > invalid image name (slax-6.0.6) > i must have done something wrong using the remove.Hmm. This would seem to be more of a cobbler issue. You could try editing the file for that image in: /var/lib/cobbler/config/images.d And then restarting the daemon again. I don't see anything wrong in that syntax. I've cc'd Mike DeHaan as he may be able to help with this. Perry
Carb, Brian A wrote:> Thanks Perry, > > I had to remove one of the files in /var/cobbler/config/systems.d that contained a definition with slax-6.0.6 as the image name. > > So now i've re-added this iso with the syntax > cobbler image add --name=slax-6.0.6 --file=192.168.50.2:/cobblernfs/slax-6.0.6.iso --image-type=iso > > but start of vm errors this time with: > libvir: error : internal error '/bin/mount -t nfs 192.168.50.2::/cobblernfs /mnt/IqifnNg3BmoFuvQM' exited with non-zero status 32 and signal 0: mount.nfs: Unknown error 521 > > If i issue the command manually with a single colon instead of double colon after the ip then the mount works. > > so now if i redo the cobbler add with --file=192.168.50.2/cobblernfs... (no colon at all) then the vm image starts!That is still not the right syntax See the following thread: https://www.redhat.com/archives/ovirt-devel/2008-December/msg00247.html This thread occurred a few days after the one I mentioned in my previous mail and in this discussion we clarify that specification of a url without a leading protocol is ambiguous and should not be allowed. A message in that thread points out: https://fedorahosted.org/cobbler/wiki/AllAboutImages Which outlines the two accepted formats for specifying images. Glad you got it working though :) But try recreating the image record using the nfs:// format and see if that works too. Perry