Apply configuration from cloud metadata.
Go to file
Zuul 14dfe905bd Merge "Update the shebang to python3" 2024-03-08 15:01:24 +00:00
os_apply_config Merge "Update the shebang to python3" 2024-03-08 15:01:24 +00:00
zuul.d Update python job template 2023-10-18 14:58:16 +09:00
.coveragerc coveragerc: Remove non-existent path 2023-12-19 21:00:16 +09:00
.gitignore Ignore .eggs directory 2023-12-19 21:04:48 +09:00
.gitreview OpenDev Migration Patch 2019-04-19 19:32:01 +00:00
.stestr.conf Switch to stestr 2018-07-10 18:32:59 +07:00
LICENSE Aligned tests with OpenStack standards. 2013-03-28 03:40:38 +01:00
README.rst Sync Sphinx requirement 2019-06-20 10:35:21 +08:00
requirements.txt Remove workaround for Python 2 2023-11-28 00:22:45 +09:00
setup.cfg Remove unnecessary egg_info options 2023-12-20 14:08:38 +09:00 Remove workaround for Python 2 2023-11-28 00:22:45 +09:00
test-requirements.txt Bump hacking 2024-01-17 01:37:23 +09:00
tox.ini Display coverage report 2024-02-07 14:17:57 +09:00


Team and repository tags



Apply configuration from cloud metadata (JSON)

What does it do?

It turns metadata from one or more JSON files like this:

{"keystone": {"database": {"host": "", "user": "keystone", "password": "foobar"}}}

into service config files like this:

connection = mysql://keystone:foobar@
...other settings...


Just pass it the path to a directory tree of templates:

sudo os-apply-config -t /home/me/my_templates

By default it will read config files according to the contents of the file /var/lib/os-collect-config/os_config_files.json. In order to remain backward compatible it will also fall back to /var/run/os-collect-config/os_config_files.json, but the fallback path is deprecated and will be removed in a later release. The main path can be changed with the command line switch --os-config-files, or the environment variable OS_CONFIG_FILES_PATH. The list can also be overridden with the environment variable OS_CONFIG_FILES. If overriding with OS_CONFIG_FILES, the paths are expected to be colon, ":", separated. Each json file referred to must have a mapping as their root structure. Keys in files mentioned later in the list will override keys in earlier files from this list. For example:

OS_CONFIG_FILES=/tmp/ec2.json:/tmp/cfn.json os-apply-config

This will read ec2.json and cfn.json, and if they have any overlapping keys, the value from cfn.json will be used. That will populate the tree for any templates found in the template path. See for a program that will automatically collect data and populate this list.

You can also override OS_CONFIG_FILES with the --metadata command line option, specifying it multiple times instead of colon separating the list.

os-apply-config will also always try to read metadata in the old legacy paths first to populate the tree. These paths can be changed with --fallback-metadata.


The template directory structure should mimic a root filesystem, and contain templates for only those files you want configured. For example:

~/my_templates$ tree
+-- etc
    +-- keystone
    |    +-- keystone.conf
    +-- mysql
          +-- mysql.conf

An example tree can be found here.

If a template is executable it will be treated as an executable template. Otherwise, it will be treated as a mustache template.

Mustache Templates

If you don't need any logic, just some string substitution, use a mustache template.

Metadata settings are accessed with dot ('.') notation:

connection = mysql://{{keystone.database.user}}:{{keystone.database.password}}@{{}}/keystone

Executable Templates

Configuration requiring logic is expressed in executable templates.

An executable template is a script which accepts configuration as a JSON string on standard in, and writes a config file to standard out.

The script should exit non-zero if it encounters a problem, so that os-apply-config knows what's up.

The output of the script will be written to the path corresponding to the executable template's path in the template tree:

#!/usr/bin/env ruby
require 'json'
params = JSON.parse
puts "connection = mysql://#{c['keystone']['database']['user']}:#{c['keystone']['database']['password']}@#{c['keystone']['database']['host']}/keystone"

You could even embed mustache in a heredoc, and use that:

#!/usr/bin/env ruby
require 'json'
require 'mustache'
params = JSON.parse

template = <<-eos
connection = mysql://{{keystone.database.user}}:{{keystone.database.password}}@{{}}/keystone


# tweak params here...

puts Mustache.render(template, params)

Quick Start

# install it
sudo pip install -U git+

# grab example templates
git clone /tmp/config

# run it
os-apply-config -t /tmp/config/elements/nova/os-apply-config/ -m /tmp/config/elements/seed-stack-config/config.json -o /tmp/config_output