d500651367
In sphinx, we have a :cgit_file: directive that makes links to files. Thing is - we're not using cgit anymore. So just rename it to git_file. Change-Id: I80aca5fb3cc84281e29843944fea33e6f4d9fe6f
142 lines
5.7 KiB
ReStructuredText
142 lines
5.7 KiB
ReStructuredText
:title: Puppet Master
|
|
|
|
.. _puppet-master:
|
|
|
|
Puppet Master
|
|
#############
|
|
|
|
The puppetmaster server is named puppetmaster for historical reasons - it
|
|
no longer runs a puppetmaster process. There is a centralized 'hiera'
|
|
database that contains secure information such as passwords. The puppetmaster
|
|
server contains all of the ansible playbooks to run puppet apply
|
|
as well as the scripts to create new servers.
|
|
|
|
At a Glance
|
|
===========
|
|
|
|
:Projects:
|
|
* https://puppetlabs.com/
|
|
:Bugs:
|
|
* https://storyboard.openstack.org/#!/project/748
|
|
* https://tickets.puppetlabs.com/
|
|
:Resources:
|
|
* `Puppet Language Reference <https://docs.puppetlabs.com/references/latest/type.html>`_
|
|
|
|
Puppet Driving Ansible Driving Puppet
|
|
-------------------------------------
|
|
|
|
In OpenStack Infra, there are ansible playbooks that drive the running of
|
|
``puppet apply`` on all of the hosts in the inventory. That process first
|
|
copies appropriate ``hiera`` data files to each host.
|
|
|
|
The cron jobs, current configuration files and more can be done with ``puppet
|
|
apply`` but first some bootstrapping needs to be done.
|
|
|
|
You want to install these from puppetlabs' apt repo. There is a script,
|
|
:git_file:`install_puppet.sh` in the root of the system-config repository that
|
|
will setup and install the puppet client. After that you must install the
|
|
ansible playbooks and hiera config (used to maintain secrets).
|
|
|
|
Ansible and Puppet 3 is known to run on Precise, Trusty, Centos 6 and Centos 7.
|
|
|
|
.. code-block:: bash
|
|
|
|
sudo su -
|
|
git clone https://opendev.org/opendev/system-config /opt/system-config
|
|
bash /opt/system-config/install_puppet.sh
|
|
bash /opt/system-config/install_modules.sh
|
|
echo $REAL_HOSTNAME > /etc/hostname
|
|
service hostname restart
|
|
puppet apply --modulepath='/opt/system-config/modules:/etc/puppet/modules' -e 'include openstack_project::puppetmaster'
|
|
|
|
Hiera uses a systemwide configuration file in ``/etc/puppet/hiera.yaml``
|
|
and this setup supports multiple configurations. The two sets of environments
|
|
that OpenStack Infrastructure uses are ``production`` and ``development``.
|
|
``production`` is the default and the environment used when nothing else is
|
|
specified.
|
|
|
|
The hiera configuration is placed by puppet apply into common.yaml in
|
|
``/etc/puppet/hieradata/production`` and ``/etc/puppet/hieradata/development``.
|
|
The values are simple key-value pairs in yaml format. The keys needed are the
|
|
keys referenced in your ``site.pp``, their values are typically obvious
|
|
(strings, lists of strings). ``/etc/puppet/hieradata/`` and below should be
|
|
owned by ``puppet:puppet`` and have mode ``0711``.
|
|
|
|
Below the ``hieradata`` directory, there should be a ``common.yaml`` file where
|
|
settings that should be available to all servers in the infrastructure go,
|
|
and then two directories full of files. The first is ``fqdn`` which should
|
|
contain a yaml file for every server in the infrastructure named
|
|
``${fqdn_of_server}.yaml``. That file has secrets that are only for that
|
|
server. Additionally, some servers can have a ``$group`` defined in
|
|
``manifests/site.pp``. There can be a correspondingly named yaml file in the
|
|
``group`` directory that contains secrets to be made available to each
|
|
server in the group.
|
|
|
|
All of the actual yaml files should have mode 0600 and be owned by root.
|
|
|
|
Adding a node
|
|
-------------
|
|
|
|
For adding a new node to your puppet master, you can either use the
|
|
``/opt/system-config/launch/launch-node.py`` script
|
|
(see :git_file:`launch/README` for full details) or bootstrap puppet manually.
|
|
|
|
For manual bootstrap, you need to run on the new server connecting
|
|
(for example, review.opendev.org) to the puppet master:
|
|
|
|
.. code-block:: bash
|
|
|
|
sudo su -
|
|
wget https://opendev.org/opendev/system-config/raw/branch/master/install_puppet.sh
|
|
bash -x install_puppet.sh
|
|
|
|
Running Puppet on Nodes
|
|
-----------------------
|
|
|
|
In OpenStack's Infrastructure, puppet runs are triggered from a cronjob
|
|
running on the puppetmaster which in turn runs a single run of puppet apply on
|
|
each host we know about.
|
|
|
|
The entry point for this process is ``/opt/system-config/run_all.sh``
|
|
|
|
There are a few sets of nodes which have their own playbooks so that they
|
|
are run in sequence before the rest of the nodes are run in parallel.
|
|
At the moment, this allows creation of git repos on the git slaves before
|
|
creation of the master repos on the gerrit server.
|
|
|
|
If an admin needs to run puppet by hand, it's a simple matter of either
|
|
logging in to the server in question and running
|
|
`puppet apply /opt/system-config/manifests/site.pp` or, on the
|
|
puppetmaster, running:
|
|
|
|
.. code-block:: bash
|
|
|
|
ansible-playbook --limit="$HOST:localhost" /opt/system-config/playbooks/remote_puppet_adhoc.yaml
|
|
|
|
as root, where `$HOST` is the host you want to run puppet on.
|
|
The `:localhost` is important as some of the plays depend on performing a task
|
|
on the localhost before continuing to the host in question, and without it in
|
|
the limit section, the tasks for the host will have undefined values.
|
|
There is also a script, `tools/kick.sh` that takes the host as an argument
|
|
and runs the above command.
|
|
|
|
Testing new puppet code can be done via `puppet apply --noop` or by
|
|
constructing a VM with a puppet install in it and just running `puppet apply`
|
|
on the code in question. This should actually make it fairly easy to test
|
|
how production works in a more self-contained manner.
|
|
|
|
|
|
Disabling Puppet on Nodes
|
|
-------------------------
|
|
|
|
In the case of needing to disable the running of puppet on a node, it's a
|
|
simple matter of adding an entry to the ansible inventory "disabled" group.
|
|
See the :ref:`disable-enable-puppet` section for more details.
|
|
|
|
Important Notes
|
|
---------------
|
|
|
|
#. Make sure the site manifest **does not** include the puppet cron job, this
|
|
conflicts with puppet master and can cause issues. The initial puppet run
|
|
that create users should be done using the puppet apply configuration above.
|