After you've installed Puppet Enterprise, optimize it for your environment by configuring and tuning Puppet Server settings as needed.
Tune the maximum number of JRuby instances
jruby_max_active_instances setting controls the maximum number of JRuby instances to allow on the Puppet Server.
The default used in PE is the number of CPUs - 1, expressed as
$::processorcount - 1. One instance is the minimum value and four instances is the maximum value. Four JRuby instances will work for most environments. However, if you're running a large number of nodes, you might need to increase the number of instances up to
$::processorcount - 1.
Increasing the JRuby instances increases the amount of RAM used by
pe-puppetserver. When you increase JRuby instances, increase the heap size. We conservatively estimate that a JRuby process uses 512MB of RAM.
.yamlfile, tuning the
jruby_max_active_instancesby setting the following code:
puppet_enterprise::master::puppetserver::jruby_max_active_instances: <number of instances>
ruby_load_path setting determines where Puppet Server finds components such as Puppet and Facter.
The default setting is located here:
$puppetserver_jruby_puppet_ruby_load_path = ['/opt/puppetlabs/puppet/lib/ruby/vendor_ruby', '/opt/puppetlabs/puppet/cache/lib'].
Note that if you change the
libdir you must also change the
max_requests_per_instance setting determines the maximum number of requests per instance of a JRuby interpreter before it is killed.
The appropriate value for this parameter depends on how busy your servers are and how much you are affected by a memory leak. By default,
max_requests_per_instance is set to 100,000 in PE.
When a JRuby interpreter is killed, all of its memory is reclaimed and it is replaced in the pool with a new interpreter. This prevents any one interpreter from consuming too much RAM, mitigating Puppet code memory leak issues and keeping Puppet Server up.
Starting a new interpreter has a performance cost, so set the parameter to get a new interpreter no more than every few hours. There are multiple interpreters running with requests balanced across them, so the lifespan of each interpreter varies.
max_requests_per_instanceby editing your Hiera default
.yamlfile with the following code:
puppet_enterprise::master::puppetserver::jruby_max_requests_per_instance: <number of requests>
Enable or disable
environment-class-cache-enabled setting specifies whether cached data is used when updating classes in the console. When
true, Puppet Server refreshes classes using file sync, improving performance.
The default value for
environment-class-cache-enabled depends on whether you use Code Manager:
With Code Manager, the default value is enabled (
true). File sync clears the cache automatically in the background, so clearing the environment cache manually isn't required when using Code Manager.
Without Code Manager, the default value is disabled (
To enable or disable the cache:
- Open the Hiera default
.yamlfile in an editor.
- Add the following line with the appropriate setting:
puppet_enterprise::master::puppetserver:: jruby_environment_class_cache_enabled: <true OR false>
Understanding the environment_timeout setting
environment_timeout setting controls how long the master caches data it loads from an environment, determining how much time passes before changes to an environment's Puppet code are reflected in its environment.
In PE, the
environment_timeout is set to
unlimited for performance. We recommend leaving this setting as is.
For more information, see Environments limitations.
Add certificates to the puppet-admin certificate whitelist
puppet-admin certificate whitelist as needed in Hiera.
.yamlfile with the following code:
puppet_enterprise::master::puppetserver::puppet_admin_certs: - 'example_cert_name'
Disable update checking
Puppet Server (pe-puppetserver) checks for updates when it starts or restarts, and every 24 hours thereafter. It transmits basic, anonymous info to our servers at Puppet, Inc. to get update information. You can optionally turn this off.
Specifically, it transmits:
Puppet Server version
Data collection timestamp
To turn off update checking:
- Disable update checking by adding the following parameter to either the PE Master node group in the node classifier or to your Hiera configuration:
- On your master node, run Puppet.
Puppet Server configuration files
At startup, Puppet Server reads all of the
.conf files in the
conf.d directory (
conf.d directory contains the following files:
|Contains authentication rules and settings for agents and API endpoint access.|
|Contains global configuration settings for Puppet Server, including logging settings.|
|Contains settings for Puppet Server metrics services.|
|Contains Puppet Server settings specific to Puppet Enterprise.|
|Contains SSL and external Certificate Authority service configuration settings.|
|(Deprecated) Contains rules for Certificate Authority services. Superseded by |
For information about Puppet Server configuration files, see Puppet Server's config files and the Related topics links below.
This file contains Puppet Server settings specific to Puppet Enterprise, with all settings wrapped in a
||Determines where JRuby looks for gems. It is also used by the
||Sets the Puppet configuration directory's path.||
||Sets the Puppet variable directory's path.||
Optional. Sets the maximum number of requests that may be queued waiting to borrow a JRuby from the pool. Once this limit is exceeded, a 503 Service Unavailable response will be returned for all new requests until the queue drops below the limit.
Note: Don't use this solution if your managed infrastructure includes a significant number of agents older than Puppet 5.3. Older agents treat a 503 response as a failure, which ends their runs, causing groups of older agents to schedule their next runs at the same time, creating a thundering herd problem.
||Optional. Sets the upper limit in seconds for the
random sleep set as a Retry-After
header on 503 responses returned when
||Sets the Puppet configuration directory's path. The