Value report

A new Value report page in the Admin section of the console estimates the amount of time reclaimed by using PE automation. The report is configurable based on your environment. See Access the value report for more information.

Spend less time waiting on puppet infrastructure commands

The puppet infrastructure commands that use plans, for example for upgrading, provisioning compilers, and regenerating certificates, are now noticeably faster due to improvements in how target nodes are verified.

Provision a replica without manually pinning the target node

You're no longer required to manually pin the target replica node to the PE Infrastructure Agent group before running puppet infrastructure provision replica. This action — which ensures that the correct catalog and PXP settings are applied to the replica node in load balanced installations — is now handled automatically by the command.

Configure environment caching

Using new environment timeout settings, you can improve Puppet Server performance by caching long-lived environments and purging short-lived environments. For example, in the PE Master node group, in the puppet_enterprise::master class, set environment_timeout_mode = from_last_used and environment_timeout = 30m to clear short-lived environments 30 minutes from when they were last used. By default, when you enable Code Manager, environment_timeout is set to unlimited, which caches all environments.

Configure the number of threads used to download modules

A new configuration parameter, download_pool_size, lets you specify the number of threads r10k uses to download modules. The default is 4, which improves deploy performance in most environments.

Configure PE-PostgreSQL autovacuum cost limit

The cost limit value used in PE-PostgreSQL autovacuum operations is now set at a more reasonable default that scales with the number of CPUs and autovacuum workers available. The setting is also now configurable using the puppet_enterprise::profile::database::autovacuum_vacuum_cost_limit parameter.

Previously, the setting was not configurable, and it used the PostgreSQL default, which could result in database tables and indexes growing continuously.

Rerun Puppet or tasks on failed nodes only

You can choose to rerun Puppet or a task only on the nodes that failed during the initial run by selecting Failed nodes on the Run again drop down.

Run plans only when required parameters are supplied

In the console, the option to commit a plan appears after you supply all required parameters. This is to prevent plan failures by accidentally running plans without required parameters.

Schedule plans in the console and API

You can use the console to schedule one time or recurring plans and view currently scheduled plans. Additionally, you can use the schedule_plan command for scheduling one-time plan runs using the POST /command/schedule_plan endpoint.

Use sensitive parameters in plans

You can use the Sensitive type for parameters in plans. Parameters marked as Sensitive aren't stored in the orchestrator's database, aren't returned via API calls, and don't appear in the orchestrator's log. Sensitive parameters are also not visible from the console.

View more details about plans

The environment a plan was run from is now displayed in the console on the Job details page. The environment is returned from the /plan_jobs endpoint using the new environment key. See GET /plan_jobs for more information.

Additionally, the parameters supplied to tasks that are run as part of a plan are displayed in the console on the Plan details page. Sensitive parameters are masked and are never stored for a task run.

Differentiate between software and driver patch types for Windows

PE now ignores Driver update types in Windows Updates by default and only includes the Software type, cutting down on unnecessary patch updates. To change this default, configure the new windows_update_criteria parameter in the pe_patch class by removing or changing the Type argument. See Patch management parameters for more information about the parameter.

Serve patching module files more efficiently

Certain pe_patch module files are now delivered to target nodes using methods that improve scalability and result in fewer file metadata checks during Puppet runs.

Receive a notification when CA certs are about to expire

The console now notifies you if a CA certificate is expiring soon. The Certificates page in the sidebar displays a yellow ! badge if a certificate expires in less than 60 days, and a red ! badge if a certificate expires in less than 30 days. If there are certificates that need signing in addition to the certificates expiring, the number of certificates that need to be signed is displayed in the badge but the color stays the same.

View details about a particular code deployment

The Code Manager /deploys/status endpoint now includes the deployment ID in the "deploys-status" section for incomplete deploys so you can correlate status output to a particular deployment request.

Additionally, you can query the Code Manager /deploys/status endpoint with a deployment ID to see details about a particular deployment. The response contains information about both the Code Manager deploy and the sync to compilers for the resulting commit.

Troubleshoot code deployments

File sync now uses the public git SHA recorded in the signature field of the .r10k-deploy.json file instead of an internal SHA that file sync created. Additionally, versioned directories used for lockless deploys now use an underscore instead of a hyphen so that paths are valid environment names. With these changes, you can now map versioned directories directly to SHAs in your control repository.

When upgrading to 2019.3 or later with versioned deploys enabled, versioned directories are recreated with underscores. You can safely remove orphaned directories with hyphens located at /opt/puppetlabs/server/data/puppetserver/filesync/client/versioned-dirs.

Report on user activities

A new GET /v2/events API that tracks more user activities, like date, time, remote IP address, user ID, and action. You can use the console to generate a report of activities on the User details page.

Platform support

This version adds support for these platforms.

Master removed from docs

Documentation for this release replaces the term master with primary server. This change is part of a company-wide effort to remove harmful terminology from our products.

For the immediate future, you’ll continue to encounter master within the product, for example in parameters, commands, and preconfigured node groups. Where documentation references these codified product elements, we’ve left the term as-is.

As a result of this update, if you’ve bookmarked or linked to specific sections of a docs page that include master in the URL, you’ll need to update your link.

Whitelist and blacklist deprecated

In the interest of removing racially insensitive terminology, the terms whitelist and blacklist are deprecated in favor of allowlist and blocklist. Classes, parameters, and file names that use these terms continue to work, but we recommend updating your classification, Hiera data, and pe.conf files as soon as possible in preparation for their removal in a future release.

These are the classes, parameters, task parameters, and file names that are affected.

Split-to-mono migration removed

The puppet infrastructure run migrate_split_to_mono command has been removed. The command migrated a split installation to a standard installation with the console and PuppetDB on the primary server. Upgrades to PE 2019.2 and later required migrating as a prerequisite to upgrade, so this command is no longer used. If you're upgrading from an earlier version of PE with a split installation, see Migrate from a split to a standard installation in the documentation for your current version.

Upgrade and puppet infrastructure commands failed if your primary server was not in the production environment

Upgrades and puppet infrastructure commands — including replica upgrade and compiler provisioning, conversion, and upgrade — failed with a Bolt::RunFailure if your primary server was not in the production environment.

This release fixes both issues, and upgrades to this version are unaffected.

Upgrade failed if a PostgreSQL repack was in progress

If a PostgreSQL repack operation was in progress when you attempted to upgrade PE, the upgrade could fail with the error cannot drop extension pg_repack because other objects depend on it.

Upgrade failed with an unenabled replica

PE upgrade failed if you had a provisioned, but not enabled, replica.

Compiler provisioning failed if a single compiler was unresponsive

puppet infrastructure commands failed with an external node classifier

With an external node classifier, puppet infrastructure commands, such as puppet infrastructure compiler upgrade and puppet infrastructure provision compiler, failed.

Automated Puppet runs could fail after running compiler or certificate regeneration commands

After provisioning compilers, converting compilers, or regenerating certificates with puppet infrastructure commands, automated Puppet runs could fail because the Puppet service hadn't restarted.

puppet infrastructure recover_configuration misreported success if specified environment didn't exist

If you specified an invalid environment when running puppet infrastructure recover_configuration, the system erroneously reported that the environment's configuration was saved.

Runs, plans, and tasks failed after promoting a replica

After promoting a replica, infrastructure nodes couldn't connect to the newly promoted primary server because the master_uris value still pointed to the old primary server.

This release fixes the issue for newly-provisioned replicas, however if you have an enabled replica, in both the PE Agent and PE Infrastructure Agent node groups, in the puppet_enterprise::profile::agent class, verify that the setting for master_uris matches the setting for server_list. Both values must include both your primary server and replica, for example ["PRIMARY.EXAMPLE.COM", "REPLICA.EXAMPLE.COM"]. Setting these values ensures that agents can continue to communicate with the promoted replica in the event of a failover.

Replica commands could leave the Puppet service disabled

The reinitialize replica command as well as the provision replica command, which includes reinitializing, left the Puppet service disabled on the replica.

Skipping agent configuration when enabling a replica deleted settings for the PE Agent group

If you used the --skip-agent-config flag with puppet infra enable replica or puppet infra provision replica --enable, any custom settings that you specified for server_list and pcp_broker_list in the PE Agent node group were deleted.

Provisioning a replica failed after regenerating the primary server certificate

If you previously regenerated the certificate for your primary server, provisioning a replica can failed due to permission issues with backed up directories.

Console was inaccessible with PE set to IPv6

If you specified IPv6, PE Java services still listened to the IPv4 localhost. This mismatch could prevent access to the console as Nginx proxied traffic to the wrong localhost.

Apply blocks failed to compile

Puppet Server might have failed to compile apply blocks for plans when there were more than eight variables, or when variables had names that conflict with key names for hashes or target settings in plans.

Yaml plans displayed all parameters as optional in the console

Yaml plans listed all parameters as having default values, regardless of whether there is a default value set in the code or not. This caused all parameters to display defaults in orchestrator APIs and show as optional in the console. Yaml plans no longer display all parameters as optional.

Running puppet query produced a cryptic error

ERROR - &{<nil>  } (*models.Error) is not supported by the TextConsumer, can be resolved by supporting TextUnmarshaler interface

Primary server reported HTTP error after Qualys scan

When running a Qualys scan, the primary server no longer reports the error "HTTP Security Header Not Detected. Issue at Port 443".

Nodes CSV export failed with PQL query

The csv export functionality no longer produces an error when you specified nodes using a PQL query.

The wait_until_available function didn’t work with multiple transports

When a target included in the TargetSpec argument to the wait_until_available plan function used the ACE (remote) transport, the function failed immediately and wouldn't wait for any of the targets in the TargetSpec argument.

Unnecessary logs and failed connections in bolt-server and ace-server

When requests were made with unsupported ciphers, bolt-server and ace-server would log stack traces. Stack traces might lead to unusual growth in the logs for those services when, for example, they are scanned by security scanning products. The Puma Server library in those services has been updated to prevent emitting the stack traces into the bolt-server.log and ace-server.log.

Patch task could misreport success for Windows nodes

When patching Windows nodes, running the pe_patch::patch_server task always reported success, even if there were problems installing one or more updates. With this fix, the task now fails with an error message about which updates couldn't be installed successfully.

The pe_patch fact didn't consider classifier environment group

When pe_patch scheduled scripts that uploaded facts, the facts didn't consider the current environment the node was compiling catalogs in. If the local agent environment didn’t match the environment specified by the server, the facts endpoint included tags for an unexpected environment.

Reenabling lockless code deploys could fail

Reenabling lockless code deploys could fail due to the persistence of the versioned code directory. With this release, any existing versioned code directory is deleted – and recreated – when you reenable lockless code deploys.

File-sync client repo grew after frequent commits

The file-sync client repo no longer grows rapidly when there are frequent commits to it. For example, when syncing the CA dir for DR, and many new certificates are signed are revoked quickly.

Value reporting

A new values API reports details about automated changes that PE makes to nodes, and provides an estimate of time freed by each type of change based on intelligent defaults or values you provide. You can also specify an average hourly salary and see an estimate of cost savings for all automated changes.

Console navigation and workflow improvements

Compiler conversion runs in parallel

When you convert all compilers at one time with puppet infrastructure run convert_legacy_compiler all=true, the process is now noticeably faster due to streamlining in when Puppet runs occur on target hosts.

Console displays enum and boolean plan parameter values in select menu

You can select plan parameters that are boolean or enum types from a drop down menu in the Value field.

Updates to metrics endpoints

Access to endpoints under /metrics are now controlled by trapperkeeper-authorization and configured in the Puppet Serverauth.conf file. The default rule allows remote access with a valid Puppet certificate.

Setting the v2 metrics endpoint to debug no longer displays debug messages from Jolokia. In order to see debugging messages, set a configuration value in addition to the usual logback changes.

Application orchestration features in the Puppet language

Upgrading Windows agents using the puppet_agent module could restart non-Puppet services

If you're using a log aggregator, upgrading Windows agents using the puppet_agent module could cause non-Puppet services to restart.

Upgrading agents using the puppet_agent module could produce non-breaking errors

Upgrading agents from versions 6.14 or 6.15 using the puppet-agent module could produce errors about an unavailable file resource or unknown HTTP resource. These errors occurred only during the initial Puppet agent run, when the agent was still using versions 6.14 or 6.15 with an updated primary server. The error resolved after the puppet-agent service restarted.

Pre-upgrade check produced inaccurate errors on standalone PE-PostgreSQL nodes

## Pre-Upgrade Checks
                    Warning: Puppet agent is not running.
                    Error: No configuration file found at /etc/puppetlabs/client-tools/services.conf. This
                    file is installed automatically on Puppet Server nodes. Make sure you are
                    running the command on a primary master, primary master replica, or compile
                    master.
                    Error: Try 'puppet infrastructure help status' for usage

The error occurred because the pre-upgrade check verified services running on the primary server which were not present on standalone PE-PostgreSQL nodes.

Upgrade could fail with custom structured facts

If you use custom facts that use structured facts, upgrade could fail with an error related to your custom fact, for example: undefined method '[]' for nil:NilClass (Puppet::Error).

Upgrade commands failed if PXP agents were configured to connect to load balancers

In installations with load balancers, the puppet infrastructure upgrade commands could fail if the PXP agent on infrastructure nodes connected to load balancers instead of to the primary server. The upgrade plan now verifies configuration and prompts you to fix any issues before continuing with the upgrade.

Compiler upgrade could fail to upgrade Puppet Server

The puppet infrastructure upgrade compiler command could fail to upgrade Puppet Server depending on how the catalog was built for performing the upgrade.

Converting all legacy compilers failed in disaster recovery installations

With disaster recovery enabled, the command to convert legacy compilers with the option all=true failed.

Converting legacy compilers failed with autosigning enabled

Running puppet infrastructure run convert_legacy_compiler with autosigning enabled caused the conversion to fail during certificate regeneration.

Converting legacy compilers could fail with DNS alternative names

If dns_alt_names were specified in the [agent] section of puppet.conf, the puppet infrastructure run convert_legacy_compiler command failed because it didn't recognize the alternative names. As a temporary workaround, we recommended moving dns_alt_names to the [main] section of puppet.conf on the compilers to be converted, however [agent] is the preferred section to specify this parameter. The compiler conversion command now recognizes DNS alternative in either the [agent] or [main] section of puppet.conf.

Missing package dependencies for SUSE Linux Enterprise Server agent nodes

On agent nodes running SUSE Linux Enterprise Server 15, the libyaml-cpp package and operating system packages prefixed with libboost_ were no longer bundled with Puppet agent, and also might not have been included in the operating system.

Command to regenerate agent certificates didn't work with nodes behind a load balancer

In large and extra-large installations with load balancers, the command puppet infrastructure run regenerate_agent_certificate failed because compilers didn't have the tasks needed to run the command, and agent nodes don't communicate directly with the primary server.

With lockless code deploy enabled, deleted branches could increase disk use

If you deleted a branch from your control repository with lockless deploys enabled, some artifacts could remain on disk and increase your disk use.

With lockless code deploy enabled, deploying with --wait could produce an erroneous timeout

Deploying code from the command line or API with the --wait flag produced a timeout error, even though the code deploy completed.

The blackout_windows parameter in pe_patch class couldn't handle time zones with negative UTC offset

If you used a negative value to offset the timezone when setting the blackout_windows parameter for patching node groups, the pe_patch fact would return an error.

The pe_patch fact wouldn't generate if there was a parsing error

The pe_patch fact couldn't be generated if there was an error when parsing the latest cached catalog for the node. Additionally, if you did not have puppetlabs-stdlib installed, packages that were fixed to a particular version in the node's catalog were not recognized by pe_patch.

Node search input didn't respond to Enter key

The node name search bar on the Nodes page in the console didn't respond to the Enter key to search for a node and you had to select Submit manually. You can now use Enter to search for nodes.

Console radiator bars had a width of zero

In the console, the colored bars in the radiator were broken and didn't show the correct layout. The radiator has been fixed.

Patch management

You can now manage patches on *nix and Windows nodes in the Patch Management section of the console. After setting up patching node groups, you can view the patch status for your nodes, filter available patches by type and operating system, and run a pre-filled task to apply patches to selected nodes from the Patches page. For information on configuring patch management and applying patches, see Managing patches.

Lockless code deploys

Using Code Manager, you can now optionally deploy code to versioned code directories rather than the live code directory. This change enables Puppet Server to continue serving catalog requests even as you deploy code.

You can enable lockless code deploys by setting puppet_enterprise::profile::master::versioned_deploys to true. For more information about lockless code deploys, see Deploy code without blocking requests.

Improvements to puppet infrastructure upgrade commands

When you specify more than one compiler to upgrade, the puppet infrastructure upgrade compiler command now upgrades all compilers at the same time, rather than sequentially. Additionally, with both the compiler and replica upgrade commands, you can now specify the location of an authentication token other than the default. For example: puppet infrastructure upgrade compiler --token-file=<PATH_TO_TOKEN>.

More secure code deploys

Permissions for the Puppet code directory are now managed by file sync directly, instead of relying on symlinks. This change improves security during code deployment.

Logging for puppet infrastructure commands that use the orchestrator

A new log file located at /var/log/puppetlabs/installer/orchestrator_info.log contains run details about puppet infrastructure commands that use the orchestrator, including the commands to provision and upgrade compilers, convert legacy compilers, and regenerate agent and compiler certificates.

Improved error handling for plans

Before running plans, the built-in check for node connectivity now provides more descriptive error messages, such as host key verification failures.

Unspecified default values for tasks and plans are supplied automatically

New scheduling options in the console

You can now specify scheduled tasks and Puppet jobs to run every two weeks or every four weeks.

Plan support for apply() on pcp transports

Plans now support using the apply_prep() function and blocks of Puppet code within calls to apply(). The feature is only available on targets connected to PE using the PCP transport and does not work on nodes connected over SSH or WinRM.

Support for new options in the command/deploy endpoint

Platform support

This version adds support for these platforms

Razor removed

Support for bolt.yaml settings in plans removed

Settings from bolt.yaml are no longer read from the environment directory. The modulepath setting is only configurable from environment.conf.

Platforms removed

Support for these platforms is removed in this release:

Upgrade removed custom classification rules from PE Master node group

Custom rules that you used to classify compilers in the PE Master node group were removed upon upgrade, or when you ran puppet infrastructure configure.

Upgrade failed with a Could not retrieve facts error

Could not retrieve facts ... undefined method `split' for nil:NilClass (Puppet::Error)from /opt/puppetlabs/installer/lib/ruby/gems/2.5.0/gems/facter-4.0.20/lib/custom_facts/util/loader.rb:125:in `load'

Upgrading a replica could temporarily lock the agent on the primary server

If you tried to run Puppet on your primary server before the puppet infrastructure upgrade replica command completed, you could encounter an error that a Puppet run was already in progress.

FIPS installs didn't fully support cert chain validation

In FIPS environments, RBAC could not connect to LDAP using a pem or jks file.

Command to remove old PostgreSQL versions failed on Ubuntu

When run on Ubuntu nodes, the puppet infrastructure run remove_old_postgresql_versions command failed, erroneously reporting that PostgreSQL wasn't installed.

Enabling a replica could fail immediately after provisioning

When running puppet infrastructure provision replica --enable, the command could fail after the replica was provisioned but before it was enabled if services on the replica were still starting up. The command now waits for services to start and verifies that replication has completed before enabling the replica.

Ubuntu 20.04 couldn't be installed with PE package management

Ubuntu 20.04 wasn't available for installation as a pe_repo class, even though it was a supported agent platform.

Loading plan lists crashed console services

When plan run results were large, the console crashed due to high memory usage on the Plan details page. An optional results query parameter has been added to the GET/plan_jobs endpoint. This parameter keeps you from experiencing high memory usage in the console when loading results for large plan runs.

Default value for tasks and plans dropped in middleware

When a task had a default value of false or null, the console metadata panel did not display the default value.

Event inspector displayed wrong table types

Browsing the event inspector sometimes created inconsistencies in tables and errors in table links.