RETIRED, Disk image elements for deployment images of OpenStack
Go to file
Ben Nemec 638a863db8 Remove mysql-dev dependency from os-svc-install
Looking through the image build logs, I noticed we were pulling in
the mysql-devel packages for some reason.  It turns out this is a
historical artifact of the old TripleO source installs using
os-svc-install.  It was universally installing mysql-python on
every image where it was used because that package wasn't in the
requirements.txt for the OpenStack projects.

This was wrong in the first place, and now that we aren't supporting
that installation method anymore it doesn't make sense to keep
installing nine development packages that we'll never use on our
images.

Note that the only thing we use from os-svc-install now appears to
be os-svc-enable, so this should have no effect on the current
TripleO flow, except slightly smaller/faster image builds.

Change-Id: Iae3e6ae3f57de57639c4259361e82c8150fe2f0e
2016-03-10 13:29:10 -06:00
doc/source PEP8 on the doc and elements files 2014-08-19 10:02:00 +02:00
elements Remove mysql-dev dependency from os-svc-install 2016-03-10 13:29:10 -06:00
tests Add unit test to enforce unique file names. 2014-06-23 12:17:50 -04:00
.gitignore Look for local.json in preference to config.json. 2014-02-14 14:27:12 +13:00
.gitreview Update stackforge references to openstack 2013-08-17 22:57:21 -04:00
.testr.conf Add a .testr.conf configuration and ignore .testrepository. 2012-12-14 20:17:39 +13:00
HACKING.rst Create HACKING.rst for elements 2014-11-21 10:35:17 +00:00
LICENSE Fix copyrights for HP work. 2012-11-15 16:20:32 +13:00
MANIFEST.in Package with pbr 2013-08-06 12:50:39 +12:00
README.md Mention dib-utils package in sample instructions 2014-11-04 14:40:16 +01:00
babel.cfg Make it possible for openstack-CI to run tests 2013-02-04 22:26:17 -08:00
requirements.txt Update pbr version to match global requirements 2015-07-02 16:53:06 +01:00
run-flake8 PEP8 on the doc and elements files 2014-08-19 10:02:00 +02:00
setup.cfg Remove hardcoded version. 2014-08-20 11:01:36 +12:00
setup.py Align build files and requirements 2013-09-28 23:18:29 -04:00
test-requirements.txt Bump hacking to 0.9.x series 2014-11-14 15:40:31 +00:00
tox.ini Deprecated tox -downloadcache option removed 2016-02-19 20:15:08 +00:00

README.md

Image building rules for OpenStack images

These elements are used to build disk images for deploying OpenStack via Heat. They are built as part of the TripleO (https://wiki.openstack.org/wiki/TripleO) umbrella project.

Instructions

Checkout this source tree and also the diskimage builder, export an ELEMENTS_PATH to add elements from this tree, and build any disk images you need.

virtualenv .
source bin/activate
pip install dib-utils pyyaml
git clone https://git.openstack.org/openstack/diskimage-builder.git
git clone https://git.openstack.org/openstack/tripleo-image-elements.git
export ELEMENTS_PATH=tripleo-image-elements/elements
diskimage-builder/bin/disk-image-create -u base vm bootstrap local-config stackuser heat-cfntools -a i386 -o bootstrap

Common element combinations

Always include heat-cfntools in images that you intend to boot via heat : if that is not done, then the user ssh keys are not reliably pulled down from the metadata server due to interactions with cloud-init.

Architecture

OpenStack images are intended to be deployed and maintained using Nova + Heat.

As such they should strive to be stateless, maintained entirely via automation.

Configuration

In a running OpenStack there are several categories of config.

  • per user - e.g. ssh key registration with nova: we repeat this sort of config every time we add a user.
  • local node - e.g. nova.conf or ovs-vsctl add-br br-ex : settings that apply individually to machines
  • inter-node - e.g. credentials on rabbitmq for a given nova compute node
  • application state - e.g. 'neutron net-create ...' : settings that apply to the whole cluster not on a per-user / per-tenant basis

We have five places we can do configuration in TripleO:

  • image build time
  • in-instance heat-driven (ORC scripts)
  • from outside via APIs
  • orchestrated by Heat

Our current heuristic for deciding where to do any particular configuration step:

  • per user config should be done from the outside via APIs, even for users like 'admin' that we know we'll have. Note that service accounts are different - they are a form of inter-node configuration.
  • local node configuration should be done via ORC driven by Heat and/or configuration management system metadata.
  • inter-node configuration should be done by working through Heat. For instance, creating a rabbit account for a nova compute node is something that Heat should arrange, though the act of creating is probably done by a script on the rabbit server - triggered by Heat - and applying the config is done on the compute node by the local node script - again triggered by Heat.
  • application state changes should be done from outside via APIs

Copyright

Copyright 2012,2013 Hewlett-Packard Development Company, L.P. Copyright (c) 2012 NTT DOCOMO, INC.

All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.