Skip to content

Latest commit

 

History

History
345 lines (237 loc) · 17.5 KB

octofacts-updater.md

File metadata and controls

345 lines (237 loc) · 17.5 KB
Raw

Configuring octofacts-updater

octofacts-updater is a command line utility that anonymizes and sanitizes facts, builds and maintains index files, and (optionally) interacts directly with the GitHub API to create pull requests if there are any changes.

If you followed the quick start tutorial, you manually obtained a fact fixture by running facter on a host. However, since you did so without setting up a configuration file, your fact fixture may have contained sensitive information (e.g. the private SSH keys for the host). You also had to SSH into a host manually to run facter which is non-ideal for an automated setup.

This document will help you address all of these security and automation non-optimalities.

Configuration file quick start

The easiest way to get started with octofacts-updater is to download and install our "quickstart" configuration file. In the git repo, this is found at /examples/config/quickstart.yaml. If you'd like to download the latest version directly from GitHub, you can use wget, like this:

wget -O octofacts-updater.yaml https://raw.githubusercontent.com/github/octofacts/master/examples/config/quickstart.yaml

Data sources

octofacts-updater can obtain facts from the following data sources:

  • Local files
  • PuppetDB
  • SSH

octofacts-updater will attempt to obtain facts from each of those data sources, in the order listed above. If retrieving facts from one data source succeeds, the subsequent data sources will not be contacted. If all of the data sources are unconfigured or fail, then an error is raised.

If you are running octofacts-updater from the command line, you can force a specific data source to be used with the --datasource option. For example:

octofacts-updater --datasource localfiles ...
octofacts-updater --datasource puppetdb ...
octofacts-updater --datasource ssh ...

Local files

As seen in the quick start tutorial, if you make the facts for a node available in a YAML file, octofacts-updater can import it. This is great for testing, but it can also be used in complex environments where you cannot easily use the other built-in capabilities. In such a case, you can generate the YAML file with facts via some other method, and then import the result into octofacts.

There is no configuration needed in the octofacts-updater configuration file for the local file data source. Simply provide the full path to the file containing the facts on the command line as follows:

octofacts-updater --datasource localfiles --config-override localfile:path=/tmp/facts.yaml --hostname <hostname> ...

PuppetDB

octofacts-updater can connect to PuppetDB (version 3.0 or higher) and retrieve the facts from the most recently reported run of Puppet on the node.

You can configure the PuppetDB connection by supplying the URL in the octofacts-updater configuration file.

puppetdb:
  url: https://puppetdb.example.net:8081

SSH

octofacts-updater can SSH to a node and run the command of your choice. There are two common strategies for this option: obtaining the facts from the cache of a puppetserver, or contacting an actual node to ask for its facts.

When configuring SSH connectivity, you must supply the following parameters:

  • server: The system to SSH to.
  • user: The user to log in as.
  • command: The command to run on the target system.

You may use %%NODE%% in either the server or command parameter. This will be replaced with the hostname you have requested via the --hostname parameter.

Under the hood, octofacts-updater uses the Ruby net-ssh gem. Any other options you supply to the ssh section will be symbolized and passed to the module. For example, you may use:

  • password: Hard-code a password, instead of using public key authentication.
  • port: Choose a port other than the default port, 22.
  • passphrase: Hard-code the passphrase for a key.

Obtaining the facts from the cache of a puppetserver

In its default installation, the puppetserver will save a copy of the most recent facts for a node in the /opt/puppetlabs/server/data/puppetserver/yaml/facts directory. You can configure octofacts-updater to SSH to the puppetserver and cat the node's file.

ssh:
  server: puppetserver.example.net
  user: puppet
  command: cat /opt/puppetlabs/server/data/puppetserver/yaml/facts/%%NODE%%.yaml

Note that the /opt/puppetlabs/server/data/puppetserver/yaml/facts may be limited, via Unix file permissions, to be accessible only to a puppet user. You may need to work around this by adding the user to an appropriate group, or by enabling the command to run under sudo without a password.

Contacting an actual node to ask for its facts

The SSH data source can be leveraged to contact the node whose facts are being determined, by using %%NODE%% as the server name. The following command will contact the node in question and run facter to get the facts in YAML format.

ssh:
  server: %%NODE%%
  user: puppet
  command: /opt/puppetlabs/puppet/bin/facter -p --yaml

Note that facter may need to run as root to gather all of the facts for the system. You may need to work around this by enabling the command to run under sudo without a password.

Also, if you are using Puppet 4 or later, but are relying on "legacy" facts that were used in Puppet 3, you may need to add --show-legacy to the facter command line.

Indexing facts

octofacts supports an index of facts, allowing you to select fixtures dynamically based on parameters, rather than by specifying the name of a specific file in your tests. When properly set up, the use of an index in rspec-puppet tests will look like this example:

let(:facts) { Octofacts.from_index(app: "my_app", role: "my_role") }

To set up the index, you need to configure three settings:

  • file: The absolute or relative path to the index file.
  • node_path: The absolute or relative path to the directory where fixtures for nodes are stored.
  • indexed_facts: An array of facts that you want to index. For the example above, you would need to index the app and role facts.

In your configuration, this might look like:

index:
  file: ../spec/fixtures/facts/octofacts-index.yaml
  node_path: ../spec/fixtures/facts/octofacts
  indexed_facts:
    - app
    - role

Once you have configured the index, you can use octofacts-updater to build the index from the node fixtures you already have.

touch spec/fixtures/facts/octofacts-index.yaml
octofacts-updater --config octofacts-updater.yaml --action reindex

ℹ️ If a file or directory path starts with . or .., the path is treated as relative to the configuration file itself. This allows you to specify locations within a Puppet code repository, without regard to where on the system octofacts-updater is actually installed. Of course, you can also use absolute paths (starting with /).

Using an External Node Classifier (ENC)

If your Puppet setup uses an external node classifier (ENC), it may supply settings in its parameters hash that are treated as top level variables by Puppet code (like facts), and can be accessed in Puppet manifests. For example, if your ENC sets a top level parameter app: application_name then in your Puppet code, $::app will equal "application_name". Furthermore, %{::app} will be interpolated in a Hiera configuration file.

To configure octofacts-updater to run your ENC when updating facts for a node, add the path to your ENC in the configuration file.

enc:
  path: /usr/local/sbin/enc.sh

Note that per conventions, all ENCs must take exactly one parameter (the hostname) and return output in YAML format. You do not need to indicate the hostname (via %%NODE%% or hard-coding) when configuring the ENC, because this is assumed.

If the ENC returns parameters at the top level, these are merged in to the facts gathered for the node. In the case that the node and the ENC both contain the same key, the key from the ENC will be used.

Anonymizing and rewriting facts

To avoid committing sensitive information into source control, and to prevent rspec-puppet tests from inadvertently contacting actual systems, octofacts-updater supports anonymizing and rewriting facts. For example, you might remove or scramble SSH keys, delete or hard-code facts like system uptime that change upon each run, or change IP addresses.

You can configure this in the octofacts-updater configuration file. The quickstart example configuration contains several examples.

octofacts-updater comes with several pre-built methods to adjust facts, and supports a plugin system to allow users to extend the functionality in their own environment. (If you write a plugin that you believe may be of general use, please check our Contributing Document and look in the plugins directory.)

To configure fact adjusting, define facts as a hash in the configuration file, with one entry per adjustment.

Common actions

Deleting facts

If a fact does not contain useful information to your Puppet code, you may choose to remove it from the fact fixtures.

The following example will delete the ec2_userdata fact:

facts:
  ec2_userdata:
    plugin: delete

Setting fact to a static value

If a fact changes frequently, you may choose to set it to a static value. This will avoid making changes to the fact fixtures each time the updater runs.

The following example will set the uptime_seconds fact to 12345, which will avoid rewriting the value of this fact each time the updater runs:

facts:
  uptime_seconds:
    plugin: set
    value: 12345

Obscuring SSH private keys

To avoid committing the SSH private key of a node (or any similar credential) into source control, you can use the randomize_long_string plugin. This will generate a string of random characters that is the same length as the original key.

The quickstart example configuration invokes this for the standard SSH facts. In the following example, all facts that match the regular expression will have their values randomized.

facts:
  ssh_keys:
    regexp: ^ssh\w+key$
    plugin: randomize_long_string

Note that it is not possible to reconstruct the original key from the random key; however, the original key is used to seed the random number generator so the random key will be consistent so long as the original key does not change.

Randomizing IP addresses

To prevent rspec-puppet tests from contacting or accessing actual hosts, you may choose to use random IP addresses to replace actual IP addresses. (Best practice would dictate that you isolate your continuous integration environment from your production environment, but this doesn't always happen...)

You can use the IP anonymization plugin to adjust ipaddress fact as follows:

facts:
  ipaddress:
    plugin: ipv4_anonymize
    subnet: "10.0.1.0/24"

The original IP address is used to seed the random number generator, so the random IP address chosen will be consistent so long as the original IP address does not change. There is a corresponding ipv6_anonymize plugin to generate random IPv6 addresses on a specified IPv6 subnet.

Handling structured or multiple facts

Structured facts are organized in a hash structure, like in this example:

ssh:
  dsa:
    fingerprints:
      sha1: SSHFP 2 1 abcdefabcdefabcdefabcdefabcdefabcdefabcd
      sha256: SSHFP 2 2 abcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcd
    key: AAAAxxxxxxxxxxxx...
  ecdsa:
    fingerprints:
      sha1: SSHFP 3 1 abcdefabcdefabcdefabcdefabcdefabcdefabcd
      sha256: SSHFP 3 2 abcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcd
    key: AAAAxxxxxxxxxxxx...
  ...

It is possible to use regular expressions to "dig" into the structure with the following format:

facts:
  ssh:
    structure:
      - regexp: .+
      - regexp: ^key$
    plugin: randomize_long_string

In the example above, the code will explore the structured fact named ssh. At the first level, it will match the regular expression .+ (which is any number of any character, i.e., every element). Then in each matching level, it will match the regular expression ^key$ (which is an exact match of the string "key"). For each match, the plugin will be executed. In the example above:

ssh:
  dsa:
    fingerprints:
      sha1: SSHFP 2 1 abcdefabcdefabcdefabcdefabcdefabcdefabcd
      sha256: SSHFP 2 2 abcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcd
    key: THIS_IS_RANDOMIZED...
  ecdsa:
    fingerprints:
      sha1: SSHFP 3 1 abcdefabcdefabcdefabcdefabcdefabcdefabcd
      sha256: SSHFP 3 2 abcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcdefabcd
    key: THIS_IS_RANDOMIZED...
  ...

Extending octofacts-updater with plugins

octofacts-updater supports a plugin architecture to add additional fact anonymizers. The methods included with the gem are also written in this plugin architecture. You can read the plugin reference documentation or browse the plugins directory.

If you find yourself needing to create additional plugins for your site-specific needs, you can include those plugins by placing entries in the octofacts-updater configuration file that reference the files where those plugins exist:

plugins:
  - /usr/local/lib/octofacts-updater/custom-plugins.rb
  - /usr/local/lib/octofacts-updater/more-plugins.rb

Each plugin is created with code in the following general structure:

# Plugin name: name_of_plugin
#
# fact  - OctofactsUpdater::Fact object of the fact being manipulated
# args  - Hash with arguments specified in the configuration file (e.g. "structure" plus any other parameters)
# facts - Hash of { fact_name, OctofactsUpdater::Fact } for every fact defined (in case one fact needs to reference another)
#
# The method should adjust `fact` by calling methods such as `.set_value`. It should NOT modify `args` or `facts`.
#
OctofactsUpdater::Plugin.register(:name_of_plugin) do |fact, args = {}, facts|
  value = fact.value
  new_value = # Your code here
  fact.set_value(new_value)
end

Automating Pull Request creation with GitHub

To configure octofacts-updater to push changes to a branch on GitHub and open pull requests, use the following configuration:

github:
  branch: octofacts-updater
  pr_subject: Automated fact fixture update for octofacts
  pr_body: A nice inspiring and helpful message to yourself
  repository: example/puppet
  commit_message: Automated fact fixture update for octofacts
  base_directory: /Users/hubot/projects/puppet
  # token: We recommend that you set ENV["OCTOKIT_TOKEN"] instead of hard-coding a token here.

Your personal access token

Head to https://github.com/settings/tokens to generate an access token.

It is possible to place the token into the configuration file like:

github:
  token: abcdefg1234567

However, we recommend that you do not do this, especially if the configuration file is going to be checked in to a source code repository. Instead, you may set the environment variable OCTOKIT_TOKEN with the contents of your token. ("octokit" is a reference to the octokit gem, which underlies the GitHub integration of octofacts-updater.)

Using the GitHub integration

  1. The GitHub integration is only available when running with --action bulk, as it is designed to push a comprehensive change set.

  2. To trigger the GitHub integration, supply --github as a command line argument when running octofacts-updater. (This will raise an error if there is no github section in the configuration, or if required parameters are missing.)

  3. The base_directory setting distinguishes between the directory paths on the system you're running on, and the repository you're committing to. As an example, consider that a user has checked out their Puppet code to /Users/hubot/projects/puppet and their octofacts-managed fact fixtures are in /Users/hubot/projects/puppet/spec/fixtures/facts/octofacts. They want to create and manage files in spec/fixtures/facts/octofacts within the repo, but don't want to create a /Users/hubot/projects/puppet directory there. As a reminder, --config-override github:base_directory=/Users/hubot/projects/puppet is available to override parameters in the configuration file.

Tips

  1. Be sure that you delete the branch on GitHub once you've merged it, so that it can be recreated from the most recent default branch the next time octofacts-updater is executed.

  2. The GitHub integration does not merge the default branch into your branch automatically. You can do this on GitHub with the "Update Branch" button.

Putting it all together

Once you've configured your data source and paths, and optionally the integration to GitHub, it's time to run octofacts-updater. Assuming you've followed our examples, your command would look like this to build the initial list of fixtures:

bin/octofacts-updater --config octofacts-updater.yaml -a bulk -l <fqdn1>,<fqdn2>,<fqdn3>,...

Thereafter, the list of nodes to index will be pulled from the index file, so you won't need to list them out each time.