Puppet Enterprise 2016.4

  1. Resolved issues in PE 2016.4.8
    1. Agent installation and upgrade failed on Solaris nodes when using wget
    2. Windows agent supports data execution prevention
  2. Resolved issues in PE 2016.4.7
    1. ActiveMQ configuration file couldn’t be repaired
    2. PXP-agent service failed to restart after upgrade
    3. ActiveMQ configuration and PE classification weren’t synched after upgrade from 2015.2.1 or earlier
    4. Agent packages for AIX, Solaris, OSX, and Windows weren’t synched correctly to compile masters
    5. Nodes with no reports generated console errors on report retrieval
    6. Catalog compilation failed with older versions of the Puppet inifile module
  3. Resolved issues in PE 2016.4.6
    1. Compile masters couldn’t retrieve certain agent packages
    2. Custom environments for preconfigured node groups were re-set on upgrade
    3. Use of r10k log level debug2 caused an “invalid configuration” error
    4. Installer hung if minimum system requirements weren’t met
    5. Console stops responding when navigating new group in separate window
  4. Resolved issues in PE 2016.4.5
    1. PE installed newer puppet-agent package
    2. Web-based installer didn’t support SLES 12 (SP1, SP2)
    3. Upgrade failed if a PE Infrastructure node group name contained unexpected capitalization
    4. PuppetDB could not store sensitive parameters
    5. Enabling PuppetDB replicas could generate duplicate reports
    6. PuppetDB filters null bytes in Resource Events fields
    7. MCollective class filtering did not work on AIX agents
    8. ActiveMQ didn’t remove stored reply queues
    9. Overview load time for deployments with a large number of facts
  5. Resolved issues in PE 2016.4.3
    1. Agent applied cached catalog during split upgrades
    2. Message for a puppet code deploy -all error was unhelpful
    3. Stack trace was shown in Code Manager log instead of timeout error message
    4. File sync failed if it found a .git directory in the live code dir
    5. Some MCollective symlinks not set correctly
  6. Resolved issues in PE 2016.4.2
    1. Large class lists caused PE console lag
    2. Opting out of data collection impacted the check_for_updates flag’s setting
  7. Node classifier did not properly update restored environments
  8. Resolved issues in PE 2016.4
    1. PE installer no longer stores temp files /tmp/ dir if noexec`
    2. Code Manager could develop a bloated cache with many short-lived branches
    3. Setting strict_variables to true caused an error
    4. Symlinks caused r10k to terminate with a non-zero exit code
    5. group-children endpoint returned error for groups with no children
    6. The Run Puppet option for individual nodes wasn’t available in large installations
    7. Underscores in node names resulted in error
    8. Count of failed login attempts did not reset after reinstating a user account
    9. Installer warned of unavailable version variables
    10. SE Linux Puppet resources couldn’t be managed in previous release
    11. Console “not authorized” screen lacked a reload option

This page describes resolved issues in the latest Puppet Enterprise (PE) release.

For more information about this release, see:

Tip: This PE release contains several components that have additional release notes. Refer to Related release notes for more information.

Resolved issues in PE 2016.4.8

Agent installation and upgrade failed on Solaris nodes when using wget

Frictionless agent installation and upgrade failed on Solaris nodes if wget is used instead of curl.

Windows agent supports data execution prevention

Windows agents have to been updated to include security enhancements for data execution prevention. The agent also corrects space layout randomization.

Resolved issues in PE 2016.4.7

ActiveMQ configuration file couldn’t be repaired

If the ActiveMQ configuration file, activemq.xml was removed or damaged, PE couldn’t replace or repair the file.

PXP-agent service failed to restart after upgrade

When upgrading from PE 2016.4.2 to PE 2016.4.5, the PXP-agent service couldn’t restart on SLES 11 nodes.

ActiveMQ configuration and PE classification weren’t synched after upgrade from 2015.2.1 or earlier

When upgrading from 2015.2.1 or earlier, where hub and spoke installations could be configured without setting activemq_hubname, ActiveMQ configuration and PE classification could get out-of-sync. This desynchronization caused hub-and-spoke to be misconfigured if activemq.xml was regenerated from a fresh base file.

With this version, if activemq_hubname is not set, hub and spoke connections are inferred from nodes with the Puppet_enterprise::Profile::Amq::Hub class.

Agent packages for AIX, Solaris, OSX, and Windows weren’t synched correctly to compile masters

Applying any of the pe_repo::platform::aix_* classes on a compile master that couldn’t reach pm.puppetlabs.com resulted in a failure.

Nodes with no reports generated console errors on report retrieval

If you clicked a node’s report list, and the node didn’t have any reports stored in PuppetDB, the console returned an error that it couldn’t retrieve the report.

Catalog compilation failed with older versions of the Puppet inifile module

Using a Puppet inifile module earlier than version 1.1.0 conflicted with the PE inifile module, resulting in catalog compilation failures.

Resolved issues in PE 2016.4.6

Compile masters couldn’t retrieve certain agent packages

Enterprise Linux and Ubuntu agent packages weren’t distributed to compile masters as expected through the pe_repo class.

Custom environments for preconfigured node groups were re-set on upgrade

If you changed the environments of preconfigured node groups, upgrading re-set the environments to production.

Use of r10k log level debug2 caused an “invalid configuration” error

In PE 2016.4.5, setting the r10k log verbosity level to debug2 resulted in a Code Manager could not be started due to invalid configuration error.

Installer hung if minimum system requirements weren’t met

The graphical installer hung when verifying the target host’s CPU, memory, and disk space if those components were below recommended thresholds.

Console stops responding when navigating new group in separate window

In some cases, the PE console would stop responding if you tried to navigate a newly created node group in a separate tab or window while the group was still open in the original tab or window.

Resolved issues in PE 2016.4.5

PE installed newer puppet-agent package

Previously, PE would install a newer puppet-agent package than expected if the Puppet Collection 1 (PC1) open source Puppet repos were set up on the system. In this release, the installer installs the specific version of the puppet-agent package that is expected.

Web-based installer didn’t support SLES 12 (SP1, SP2)

Previously, the web-based installer wasn’t supported on SLES 12 (SP1 or SP2), so you had to use the text-mode installation method. Web-based installation is now supported.

Upgrade failed if a PE Infrastructure node group name contained unexpected capitalization

Previously, installation failed if you changed the default capitalization for any PE Infrastructure node group names, because the installer was case insensitive. For example, if the installer encountered a PE Mcollective (lowercase “c”) node group when it expected a PE MCollective (uppercase “c”) node group, it returned an error about non-unique group names.

PuppetDB could not store sensitive parameters

This release corrects an issue in which adding the sensitive datatype to the parameters of an existing resource caused catalog storage failures in PuppetDB.

Enabling PuppetDB replicas could generate duplicate reports

In previous versions, enabling PuppetDB replication could generate the following duplicate reports error:

ERROR: duplicate key value violates unique constraint "reports_hash_expr_idx"

These duplicates didn’t affect the storage of report data but did cause performance impacts and increased disk usage due to discarded duplicates.

PuppetDB filters null bytes in Resource Events fields

Previously, PuppetDB was unable to properly handle entries in the Windows registry that contained a default value of '\0'.

MCollective class filtering did not work on AIX agents

This release corrects an issue in which MCollective class filtering did not work on AIX agents.

ActiveMQ didn’t remove stored reply queues

In previous versions, ActiveMQ indefinitely stored queues used for MCollective replies. Each time a MCollective command was run, a small amount of memory leak would occur. This has been resolved in this release.

Overview load time for deployments with a large number of facts

Previously, the Overview page in the PE console loaded very slowly for some deployments with a large number of facts. Page load time has been significantly decreased in this version.

Resolved issues in PE 2016.4.3

Agent applied cached catalog during split upgrades

Previously, during upgrades of split installations, an agent could attempt to get a catalog but fail, and then apply a cached catalog and interrupt the upgrade.

Message for a puppet code deploy -all error was unhelpful

When running puppet code deploy --all, if a collect-failed error occured, the error message did not have information about the cause of the error, showing details: null instead. Now, if there is further information about the error, it is included in the details key in the JSON error response.

Stack trace was shown in Code Manager log instead of timeout error message

When running puppet code deploy --all, if a timeout error occurred, the Code Manager log displayed a stack trace instead of a helpful error message.

File sync failed if it found a .git directory in the live code dir

On upgrades from r10k to Code Manager, file sync failed if it encountered any .git directories, which r10k creates, in the live code directory. With this fix, you no longer need to manually remove .git directories you upgrade from r10k to Code Manager.

This release resolves an issue in which some MCollective symlink paths were not set correctly.

Resolved issues in PE 2016.4.2

Large class lists caused PE console lag

Previously in PE, if you had a large number of classes and environments and tried to access the Node Manager in the PE console, you could experience browser lags or freezes.

Opting out of data collection impacted the check_for_updates flag’s setting

The check_for_updates flag was linked to the send_analytics_data flag in Puppet Server, such that the value set for send_analytics_data also automatically applied to check_for_updates, and vice versa.

Node classifier did not properly update restored environments

In some cases, the node classifier didn’t correctly update environments that were deleted and then restored, if the environment contained the same classes. This issue most often affected Code Manager users.

Resolved issues in PE 2016.4

PE installer no longer stores temp files /tmp/ dir if noexec`

In previous versions of PE, during installation, the installer would store some temporary files in the /tmp directory. If that directory was not executable, the install would fail. The installer now stores these temporary files in /opt/puppetlabs/server/data/puppetserver/server_data if the /tmp directory is not executable.

Code Manager could develop a bloated cache with many short-lived branches

Code Manager did not support repository pruning. If you used many short-lived branches, Code Manager’s cache could become bloated and take up a lot of disk space. Code Manager now performs garbage collection on cached repositories periodically to avoid this issue.

Setting strict_variables to true caused an error

There was a disused parameter in the puppet_enterprise::profile::master class that caused an error if the Puppet setting strict_variables was set to true. The parameter has now been removed.

Previously, if r10k encountered a symbolic link when deploying a Forge module to an environment, r10k logged an error, and then when the process finished, terminated with a non-zero exit code. It now logs a warning and exits 0 in this scenario.

group-children endpoint returned error for groups with no children

Calling the group-children endpoint for a group with no children resulted in a server error, rather than the expected empty array.

The Run Puppet option for individual nodes wasn’t available in large installations

In installations with large numbers of nodes, the Run Puppet button didn’t appear on the node detail pages.

Underscores in node names resulted in error

When loading the node detail page for a node with an underscore in its name, an error appeared stating, “Error retrieving node agent list: Server Error.” Additionally, the option to run Puppet didn’t appear.

Count of failed login attempts did not reset after reinstating a user account

In cases where a user’s account was revoked due to too many incorrect login attempts, and the account was reinstated via the API by updating the is_revoked flag, the failed_login_count was not correctly reset, so the user’s next incorrect login attempt re-revoked their account.

Installer warned of unavailable version variables

This release corrects an issue in which the PE installer displayed errors that the pe_version and pe_server_version variables were undefined.

SE Linux Puppet resources couldn’t be managed in previous release

Previously, in PE, if you were running RHEL 6 or 7 agents, you couldn’t manage SE Linux Puppet resources with Puppet agent.

Console “not authorized” screen lacked a reload option

When a user lacking console access permission logged into the console, the “You are not authorized to access the console” screen did not facilitate logging out. A link to log out and return to the login screen has been added.

Back to top