openstack-nova-compute failed to start when using qemu2.0 on centos

[Moshe Levi]

after installing all the nodes we see this error in the compute nodes:
in nova-compute
2014-07-06 12:42:43.774 4304 TRACE nova.openstack.common.threadgroup rv = meth(*args,**kwargs)
2014-07-06 12:42:43.774 4304 TRACE nova.openstack.common.threadgroup File "/usr/lib64/python2.6/site-packages/libvirt.py", line 3524, in getVersion
2014-07-06 12:42:43.774 4304 TRACE nova.openstack.common.threadgroup if ret == -1: raise libvirtError ('virConnectGetVersion() failed', conn=self)
2014-07-06 12:42:43.774 4304 TRACE nova.openstack.common.threadgroup libvirtError: internal error: Cannot find suitable emulator for x86_64

and in the puppet log :
Augeas[sysconfig-libvirt](provider=augeas) (warning): Loading failed for one or more files, see debug for /augeas//error output

after investigating it with Vitaly Parakhin, /usr/bin/qemu-system-x86_64 was missing and this configuration was missing:
virt_type = kvm in nova.conf
 modprobe -a kvm
 modprobe -a kvm-intel

was build with latest commit in fuel-library:
commit 8821051a1f16340d52b2659a0f2ddf2be4274e41
and with external repository of qemu2.0 and libvirt
"qemu2,http://fuel-repository.mirantis.com/repos/centos-fuel-5.1-stable-15943/centos" "libvert,http://fuel-repository.mirantis.com/repos/centos-fuel-5.1-stable-17019/centos"

[Michael Semenov (msemenov)]

Fixed by https://review.openstack.org/#/c/105863/

[Moshe Levi (moshele)]

Hi,

I tested the fix (https://review.openstack.org/#/c/105863/ only fix) and it only solve the problem that the /usr/bin/qemu-system-x86_64 that was missing

but still the kvm kernel modules are not loaded

[Andrey Danin (gcon-monolake)]

The problem was reproduced with Public ISO #44 Jul 27, 2014
https://fuel-jenkins.mirantis.com/view/ISO/job/publish_fuel_community_iso/17/
After the deployment of Compute nodes finished, a user have to manually run 'modprobe -a kvm; modprobe -a kvm-intel; service libvirtd restart;' on each Compute node.

Can you check, is this related to package or to Puppet manifests?

[Moshe Levi (moshele)]

how can I tell?

[Andrey Danin (gcon-monolake)]

Aleksander, can you check, is this a wrong package or not? Who should do modprobe?

[Andrey Danin (gcon-monolake)]

I checked our manifests - we do nothing with modprobe kvm. It should be a package issue.

[Moshe Levi (moshele)]

we installed this http://seed.fuel-infra.org/fuelweb-iso/fuel-community-master-47-2014-07-30_05-27-27.iso.torrent iso without mellanox feature and still the kvm modules are not loaded.

I think you should see this issue in your lab as well.

[Alexander Podrepniy (apodrepniy)]

I have the same situation on build #47.

On other old build with qemu v1.2 in package qemu-kvm I see file /etc/sysconfig/modules/kvm.modules .
In build #47 this file not exists.

File /etc/sysconfig/modules/kvm.modules from build 311:
====================== cut ======================
#!/bin/sh

case $(uname -m) in
    i?86|x86_64)
        grep vmx /proc/cpuinfo >/dev/null 2>&1 && vmx=1
        grep svm /proc/cpuinfo >/dev/null 2>&1 && svm=1

        if [ "$vmx" ] || [ "$svm" ]; then
            [ "$vmx" ] && modprobe -b kvm-intel >/dev/null 2>&1 && kvm=1
            [ "$svm" ] && modprobe -b kvm-amd >/dev/null 2>&1 && kvm=1
        fi
        ;;
    ppc64)
        grep OPAL /proc/cpuinfo >/dev/null 2>&1 && opal=1

        if [ "$opal" ]; then
            modprobe -b kvm >/dev/null 2>&1 && kvm=1
        fi
        ;;
    s390x)
        modprobe -b kvm >/dev/null 2>&1 && kvm=1
        ;;
esac

if [ "$kvm" ]; then
    # no reason why vhost should not work with TCG, but currently it does not
    modprobe -b vhost-net >/dev/null 2>&1
else
    exit 0
fi
====================== cut ======================

[Kostiantyn Danylov (kdanylov)]

hacky fix
http://gerrit.mirantis.com/19859

[OSCI Robot (oscirobot)]

Package qemu has been built from changeset: http://gerrit.mirantis.com/19859
RPM Repository URL: http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable-19859/centos
You can build an ISO with this package:
make iso EXTRA_RPM_REPOS="osci-testing,http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable-19859/centos"

[Vladimir Kuklin (vkuklin)]

Guys, I think that hacky workaround is not what we want here.

1) I could not reproduce this issue on our master ISO. Let's check if it is reproducible now on ISO # 52
2) If it is confirmed, let's just use the same kvm.modules script as it seems ok.

[OSCI Robot (oscirobot)]

Package qemu has been built from changeset: http://gerrit.mirantis.com/19859
RPM Repository URL: http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable-19859/centos
You can build an ISO with this package:
make iso EXTRA_RPM_REPOS="osci-testing,http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable-19859/centos"

[OSCI Robot (oscirobot)]

Package qemu has been built from changeset: http://gerrit.mirantis.com/19859
RPM Repository URL: http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable/centos
You can build an ISO with this package:
make iso EXTRA_RPM_REPOS="osci-testing,http://osci-obs.vm.mirantis.net:82/centos-fuel-5.1-stable/centos"

[Moshe Levi (moshele)]

the fix is working