Puppet's Services: Puppet Server
This version of Puppet Server is not included in Puppet Enterprise. The latest version of PE includes Puppet Server 2.7. A newer version is available; see the version menu above for details.
Puppet master is a Ruby application that compiles configurations for any number of Puppet agent nodes, using Puppet code and various other data sources. (For more info, see Overview of Puppet’s Architecture.)
Puppet Server is an application that runs on the Java Virtual Machine (JVM) and provides the same services as the classic Puppet master application. It mostly does this by running the existing Puppet master code in several JRuby interpreters, but it replaces some parts of the classic application with new services written in Clojure.
Puppet Server is one of two recommended ways to run the Puppet master service; the other is a Rack server. Today they’re mostly equivalent — Puppet Server is easier to set up and performs better under heavy loads, but they provide the same services. In the future, Puppet Server’s features will further surpass the Rack Puppet master, and we plan to eventually disable Rack support.
Note: Puppet Enterprise 3.7 and later use Puppet Server by default. You do not need to manually install or configure it.
This page describes the generic requirements and run environment for Puppet Server; for practical instructions, see the docs for installing and configuring it. For details about invoking the
puppet master command, see the
puppet master man page.
Puppet provides Puppet Server packages for Red Hat Enterprise Linux, RHEL-derived distros, Fedora, Debian, and Ubuntu.
If we don’t provide a package for your system, you can run Puppet Server from source on any POSIX server with JDK 1.7 or later. See Running from Source for more details.
Note that Puppet Server is versioned separately from Puppet itself. Puppet Server 1.0 is compatible with Puppet 3.7.3 and later, but will not be compatible with Puppet 4.0; there will be a separate Puppet Server release to coincide with the next major version of Puppet.
Controlling the Service
The Puppet Server service name is
puppetserver. To start and stop the service, you’ll use the usual commands for your OS, such as
service puppetserver restart,
service puppetserver status, etc.
Note that while Puppet Server has better performance overall than a Ruby Puppet master running on the Apache/Passenger stack, it’s slower to start up. Starting or restarting the service will take longer than you are used to.
Puppet Server’s Run Environment
Puppet Server consists of several related services that share state and route requests among themselves. These services run inside a single JVM process, using the Trapperkeeper service framework.
From a user’s perspective, it mostly acts like a single monolithic service. (Unlike a Rack-based Puppet master, which often looks like a web server, a plugin to manage Rack apps, and a bunch of spawned Ruby processes with varying lifespans.) Most of the architectural complexity is wrapped and hidden; the main exception is the handful of extra config files that manage different internal services.
Embedded Web Server
Puppet Server uses a Jetty-based web server embedded in the service’s JVM process. Unlike with a Rack-based Puppet master, you don’t need to do anything special to configure or enable the web server; it works out of the box. Unlike the WEBrick server embedded in the Ruby Puppet master, it performs well under production-level loads.
Puppet Master Service
Puppet Server includes a “master” service that provides the same basic functions as Rack and WEBrick masters, using the same input and output formats. See Puppet V3 HTTP API for more information on the basic APIs. Puppet Server’s master service provides some additional APIs that the Rack and WEBrick Puppet masters do not.
For docs on the Puppet Server-specific APIs hosted by the master service, see:
Certificate Authority Service
Puppet Server includes a certificate authority (CA) service that accepts certificate signing requests (CSRs) from nodes, serves certificates and a certificate revocation list (CRL) to nodes, and optionally accepts commands to sign or revoke certificates. It provides these services at the same URLs used by Rack and WEBrick Puppet masters, using the same input and output formats. (The specific endpoints are
See CA V1 HTTP API for more information on these APIs.
The CA service uses .pem files in the standard Puppet
ssldir to store credentials. You can use the standard
puppet cert command to interact with these credentials, including listing, signing, and revoking certificates.
Admin API Service
Puppet Server includes an administrative API for triggering maintenance tasks. This is a new feature that doesn’t exist in Rack or WEBrick Puppet masters.
Right now, the main administrative task is forcing expiration of all environment caches. This lets you deploy new code to long-timeout environments without having to do a lengthy full restart of the service.
- For API docs, see:
- For details about environment caching, see the page about environment limitations.
Most of Puppet Server’s work — compiling catalogs, receiving reports, etc. — is still done by the Ruby-based Puppet master application. But instead of using the operating system’s MRI Ruby runtime, Puppet Server runs Puppet in JRuby, drawing from a separate collection of Ruby code.
Since we don’t use the system Ruby, you can’t use the system
gem command to install Ruby Gems for use by the Puppet master. Instead, Puppet Server includes a separate
puppetserver gem command for installing any libraries that your Puppet extensions might require. See the “Using Ruby Gems” page for details.
Additionally, if you need to test or debug code that will be used by Puppet Server, we include
puppetserver ruby and
puppetserver irb commands that will execute Ruby code in a JRuby environment identical to what the Puppet master application uses.
To handle parallel requests from agent nodes, Puppet Server maintains several separate JRuby interpreters, all independently running Puppet’s application code, and distributes agent requests among them. Today, agent requests are distributed more or less randomly, without regard to their environment; this may change in the future.
You can configure the JRuby interpreters in the
jruby-puppet section of the
You can maximize Puppet Server’s performance by tuning your JRuby configuration. To learn more about tuning your configuration, see our Puppet Server Tuning Guide.
Puppet Server needs to run as the user
pe-puppet if you are running Puppet Enterprise, or
puppet if you are running open source Puppet.
The user is specified in
/etc/sysconfig/pe-puppetserver for PE, or in
/etc/sysconfig/puppetserver for open source Puppet. (Puppet Server ignores the
group settings from puppet.conf.)
All of the Puppet master’s files and directories must be readable and writable by this user.
By default, Puppet’s HTTPS traffic uses port 8140. The OS and firewall must allow Puppet Server’s JVM process to accept incoming connections on this port.
You can change the port in
webserver.conf if necessary. See the Configuration page for details.
Puppet Server completely ignores the
masterport setting in the puppet.conf file.
All of Puppet Server’s logging is routed through the JVM Logback library. By default, it logs to
/var/log/puppetlabs/puppetserver/puppetserver.log. The default log level is ‘INFO’. By default, Puppet Server sends nothing to syslog.
All log messages follow the same path, including HTTP traffic, catalog compilation, certificate processing, and all other parts of Puppet Server’s work. This differs from Rack and WEBrick masters, which split their HTTP and application logs.
Puppet Server relies on
logrotate to manage the log file, and installs a configuration file at
Logback is heavily configurable; if you need something more specialized than a unified log file, you can probably get it. See the configuration docs for info on configuring logging.
Finally, there’s a special “daemon” log file used only for errors that happen before logging is set up or which cause the logging system to die. This file can be found at
/var/log/puppetlabs/puppetserver/puppetserver-daemon.log on platforms using SysV-style init, and in journalctl on platforms using systemd.
By default, Puppet Server handles SSL termination automatically.
In network configurations that require external SSL termination (e.g. with a hardware load balancer), you’ll need to configure a few other things. See the External SSL Termination page for details. In summary, you’ll need to:
- Configure Puppet Server to use HTTP instead of HTTPS
- Configure Puppet Server to accept SSL information via insecure HTTP headers
- Secure your network so that Puppet Server cannot be directly reached by any untrusted clients
- Configure your SSL terminating proxy to set the following HTTP headers:
X-Client-DN(mandatory for client-verified requests)
X-Client-Cert(optional; required for trusted facts)
Configuring Puppet Server
Puppet Server uses a combination of Puppet’s usual config files along with its own configuration files, which are located in the
conf.d directory contains:
global.conf: Global configuration settings for Puppet Server.
web-routes.conf: Web server configuration settings.
puppetserver.conf: Settings for Puppet Server itself, including the JRuby interpreter and the administrative API.
auth.conf: Authentication rules for Puppet Server endpoints.
master.conf(deprecated): Settings for the Puppet master functionality of Puppet Server.
ca.conf(deprecated): Settings for the Certificate Authority service.
For detailed information about Puppet Server settings and the
conf.d directory, refer to the Configuration page.
While Puppet Server can use Puppet’s
auth.conf for access control, this method is deprecated in favor of a new authentication system introduced in Puppet Server 2.2 and configured through its own
As mentioned above, Puppet Server also uses Puppet’s usual config files, including most of the settings in
puppet.conf. However, Puppet Server treats some
puppet.conf settings differently, and you should be aware of these differences.