Attempt to install Nova-Docker driver on Fedora 21

asked 2015-01-29 09:40:18 -0600

dbaxps gravatar image

updated 2015-01-31 03:23:06 -0600

UPDATE as of 01/31/2015 nova-compute.log

4b5f8d9] APIError: 500 Server Error: Internal Server Error ("Cannot start container 1d2ff4e5626d6ffaeae6690610868bfe8b299a758c2b1c08c1d0a9dfd3c6e739: write /sys/fs/cgroup/memory/system.slice/docker-1d2ff4e5626d6ffaeae6690610868bfe8b299a758c2b1c08c1d0a9dfd3c6e739.scope/memory.memsw.limit_in_bytes: invalid argument")

Per general error affecting docker on F21 , not related to nova-docker driver.

Actually attempt to reproduce
Set up Two Node RDO Juno ML2&OVS&VXLAN Cluster runnig Docker Hypervisor on Controller and KVM on Compute (CentOS 7)
After running on Controller :-

# yum install docker-io -y
# yum install -y python-pip git  
# git clone
# cd nova-docker
# git checkout stable/juno
# python install

Get popping up messages
Problem in package


Attempt to run openstack-status reports 'module' object has no attribute 'PY2' for nova service-list, keystone user-list, neutron net-list.
Fixed with :

[root@juno01svc nova-docker]# pip install --upgrade six

Downloading/unpacking six from
  Downloading six-1.9.0-py2.py3-none-any.whl
Installing collected packages: six
  Found existing installation: six 1.7.3
    Uninstalling six:
      Successfully uninstalled six
Successfully installed six
Cleaning up...

After six upgrade RDO Juno Controller&&Network and Compute nodes came back to normal work with KVM Hypervisor on Compute. Set up openstack-nova-docker service on Controller following
procedure well tested on CentOS 7.
Regarding six version on F21: originally 1.7.3, right after first python install drops to 1.2 , what causes issue described above. Final check shows 1.9.0. Notice , that on CentOS 7 python install seems not to touch six version in stays 1.7.3 before and after install.
Try to load Nova-Docker container on Controller:-

nova boot --image "boris/docker-glassfish41:latest" --flavor m1.small --key-name oskey57  --availability-zone nova:juno01svc.localdomain  --nic net-id=fe7920ec-009d-4c4d-af1f-7c3060220359 GlassFish41

Creates instance in ERROR state and doesn't create container at all ( all this works fine on CentOS 7)
As of 01/11/2015 same procedure worked OK on Fedora 21
I have activated debug=True restarted openstack services , I cannot find any ERROR messages in nova and neutron logs. Regardless I can launch VMs on Compute (KVM Hypervisor)

[root@juno02svc ~]# service libvirtd status
Redirecting to /bin/systemctl status  libvirtd.service
● libvirtd.service - Virtualization daemon
   Loaded: loaded (/usr/lib/systemd/system/libvirtd.service; enabled)
   Active: active (running) since Thu 2015-01-29 21:23:52 MSK; 31min ago
     Docs: man:libvirtd(8)
 Main PID: 1025 (libvirtd)
   CGroup: /system.slice/libvirtd.service
           └─1025 /usr/sbin/libvirtd --listen

Jan 29 21:52:10 juno02svc.localdomain libvirtd[1025]: libvirt version:, package: 2.f...g)
Jan 29 21:52:10 juno02svc.localdomain libvirtd[1025]: metadata not found: Requested metadata...nt
Jan 29 21:54:06 juno02svc.localdomain libvirtd[1025]: End of file while reading ...
edit retag flag offensive close merge delete

1 answer

Sort by » oldest newest most voted

answered 2015-01-31 03:40:16 -0600

dbaxps gravatar image

updated 2015-03-24 02:05:27 -0600

Update as of 03/24/2015. View also
Setup the most recent Nova Docker Driver via Devstack on F21
END Update

The reason seems to be recent systemd upgrade (01/21/2015) .
I got Nova-Docker Driver working ( actually Docker itself) on Fedora 21 , after :-

yum downgrade  systemd-python3 systemd-compat-libs systemd-libs systemd-python systemd libgudev1

Currently installed (working shape )

[root@juno nova(keystone_admin)]# rpm -qa | grep systemd

Another option ( more attractive for me)
Download systemd-218-3.fc22.src.rpm && install packages required for rpmbuild

$ sudo yum install audit-libs-devel autoconf  automake cryptsetup-devel \
    dbus-devel docbook-style-xsl elfutils-devel  \
    glib2-devel  gnutls-devel  gobject-introspection-devel \
    gperf     gtk-doc intltool kmod-devel libacl-devel \
    libblkid-devel     libcap-devel libcurl-devel libgcrypt-devel \
    libidn-devel libmicrohttpd-devel libmount-devel libseccomp-devel \
    libselinux-devel libtool pam-devel python3-devel python3-lxml \
    qrencode-devel  python2-devel  xz-devel

Build new systemd && perform systemd upgrade (216-17=>218-3) :-

rpm -iv systemd-218-3.fc22.src.rpm
cd rpmbuild/SPEC
rpmbuild -bb systemd.spec
cd ../RPMS/x*
yum install libgudev1-218-3.fc21.x86_64.rpm \
libgudev1-devel-218-3.fc21.x86_64.rpm \
systemd-218-3.fc21.x86_64.rpm \
systemd-compat-libs-218-3.fc21.x86_64.rpm \
systemd-debuginfo-218-3.fc21.x86_64.rpm \
systemd-devel-218-3.fc21.x86_64.rpm \
systemd-journal-gateway-218-3.fc21.x86_64.rpm \
systemd-libs-218-3.fc21.x86_64.rpm \
systemd-python-218-3.fc21.x86_64.rpm \

Even on F21 running systemd 218-3 during nova-docker driver set up , right after python install six.__version__ will be dropped to 1.2. It will result a banch of pop-ups like "problem in nova package ..."
Just respond pip install --upgrade six and proceed as normal.

edit flag offensive delete link more


You should be able to simply yum --enablerepo rawhide install systemd to get the same version of systemd without needing to rebuild the RPMs locally.

larsks gravatar imagelarsks ( 2015-02-01 07:55:02 -0600 )edit

Thank you, I already did. ;)

dbaxps gravatar imagedbaxps ( 2015-02-01 09:15:20 -0600 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools



Asked: 2015-01-29 09:40:18 -0600

Seen: 786 times

Last updated: Mar 24 '15