| Age | Commit message (Collapse) | Author |
|---|
|
To facilitate the creation of a simple OpenStack configuration with
a single control and compute node, several things should be known
at build time (since in this simple configuration, we are not using
dhcp, or other name resolution techniques):
- The IP of control node
- The IP of the compute node
- The IP of the node being built
From these values, the OpenStack components and support applications
(databases, access control, etc) are configured, as well as simple name
resolution generated at build time.
A single "hosts" bbclass should be provided with the following values:
COMPUTE_IP ?= "192.168.7.4"
COMPUTE_HOST ?= "compute"
CONTROLLER_IP ?= "192.168.7.2"
CONTROLLER_HOST ?= "controller"
MY_IP ?= "${CONTROLLER_IP}"
MY_HOST ?= "${CONTROLLER_HOST}"
The above example is for a control node, using the runqemu default addresses.
The openstack-base.bbclass is responsible for generating /etc/hosts and
/etc/hostname. Any image type that requires these values at boot tiem, should
inherit this class to allow its rootfs post population hooks to run and
generate the required configuration.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
When included in a build, this layer modifies settings to allow a controller
and compute node boot and runtime under qemu system emulation.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Creating a layer than when added to a build, modifies configuration
via appends and classes to create a configured control node.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Add a nested layer that contains the configuration and appends necessary
to deploy a compute node.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
On some targets postgresql's server processes start slowly. If they haven't
started and the admin account or other operations are attempted, they fail
with a message about not being able to communicate to the local server.
If postgres is not properly setup, then subsequent components will also fail,
since they either cannot talk to the server, or can't use the 'admin' account.
To fix this issue, we add additional sleep states, and attempt to create the
admin role 10 times, with a delay between each attempt. If we fail to contact
the server after 10 attempts, a clear message is displayed and the postinst
returns a failing return code.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
/home/root was hardcoded as the location for root's .bashrc. With this change
we'll update .bashrc in /home/root or /root.
We also add the $D prefix to the paths, to ensure there is no chance of clobbering
the build machines /root/.bashrc.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
nova controller's configuration files in /etc were being improperly
packaged in nova-controller-misc.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
In cases where the postgresql database wasn't properly started, keystone's
first run setup fails without clear error messages. With this change, we
insert longer sleep periods after attempting to start the database server,
which improves the chances of a proper startup.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
python-pycrypto must inherit autotools to have the proper parameters
passed to its configuration phase. But this same package doesn't have
a "make install" target, so the autotools added install rule errors
when run against the package.
To fix this, we pick and chose what we execute by implementing our
own do_install and calling the class routines explicitly.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
|
|
To prevent build errors to the order of evaluation of FILES_* we re-order
and clarify the contents of each package.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
|
|
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
|
|
By setting the FILES_ packaging variables via .inc local vars,
we can run into trouble with variable resolution timing and hence
packaging fails.
We temporarily switch back to explicit (and duplicated) file lists
to fix packaging
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Two fixes in this change, one is to clearly set capabilities and
trigger configure, to avoid build and configuration issues.
The second is to workaround an autoconf bug by exporting: ac_cv_func_malloc_0_nonnull=yes
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Without this fix, the fetcher may think there are multiple SRCREVs
in the SRC_URI and demand a SRCREV_FORMAT, which we don't need.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
The lib/wx subcomponent of erlang generates the following QA error:
ERROR: This autoconf log indicates errors, it looked at host include and/or library paths while determining system capabilities.
Stop configure looking for files on the host, specifically unsafe paths
such as /usr/local/include.
Additionally we should be using ${libdir} and not "lib", since 64 bit builds
will use lib64 and by using ${libdir} we adapt automatically.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
To allow unique configuration of nova for compute and controller nodes,
the nova class is split into two, but packaged largely the same way.
The compute and controller classes are introduced to hold configuration
values and operations that are used by the common packaging routines to
customize and deploy.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Sometimes it is desirable to double check the identities, roles and
ids in the system. To re-intialize the database, we can allow identity.sh
to be re run.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Updating cinder/glanc/swift/horizon/nova/keystone/quantum to the
2013.1.3 bugfix release.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Rather than support all of the hypervisors and containers that the the
default libvirt configuration adds, we specify our own KVM/lxc and
related options for a basic openstack configuration.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
To avoid conflicing installs of site.py, we can safely remove it from this
package.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Several python recipes install a site.py file. To avoid package install problems
with the conflicting file, we remove site.py from this package.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
ftp.dlitz.net is not always avaialble, switching the upstream to
pypi.python.org results in reliable builds.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
pg-config is not provided as a native tool, so if called, the host version
will be found. Since the values that are typically returned by pg-config
can be determined in our build environment, we can safely remove the calls
completely.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
Create the default openrc file for the keystone python install.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
|
|
The kernel module is needed by ebtables so the RRECOMMENDS
should be in this recipe.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
The kernel module is needed by iproute2 so the RRECOMMENDS
should be in this recipe.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
This package contains the cinder-scheduler executable and an
initscript that starts the cinder-scheduler at boot time.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
Cinder-volume requires the cinder-volumes volume group created.
The backing file for this volume is created in the postinstall
with a size of 2G on the first boot and the volume group is
created at boot time in the cinder-volume initscript.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
For the openvswitch plugin to work the br-int bridge has to
be created. To create the bridge the ovsdb-server is started
using the openvswitch-switch initscript.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
The initscript was not ran because the update-rc.d class
was not inherited.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
|
ovs_quantum is the default name for the quantum DB.
Signed-off-by: Mihai Prica <prica.mihai@gmail.com>
|
Bruce Ashfield
Mark Hatle
Mihai Prica