Puppet Enterprise 2019.0

These are the issues resolved in PE 2019.0.

The PE uninstaller now purges only /opt/puppetlabs/puppet and /opt/puppetlabs/server, leaving behind /opt/puppetlabs if there are any other files remaining in this directory. Leaving this directory in place, as necessary, prevents inadvertently removing other Puppet products, such as Bolt.

Pre-upgrade checks timed out in large installations (2019.0.2)

The upgrader checks for existing MCollective agents and halts upgrade if any are found. In large installations (tens of thousands of nodes) the MCollective check could exceed the default 30-second server_url_timeout allowed by PuppetDB, effectively blocking the upgrade.

Script installation failed on Solaris 11.4 agents (2019.0.2)

Due to a change in operating system identification for Solaris 11.4, agent installation using the install script failed with an error indicating the installation method wasn't supported.

Master and replica weren't pinned to appropriate node groups after promotion (2019.0.2)

When a replica was promoted, the retired master remained pinned to the PE Database node group, and the newly promoted master was not pinned to the group. Consequently, the database was not automatically managed after you promoted a replica.

Package updates failed on promoted replicas (2019.0.2)

On promoted HA replicas, the managed package repository configuration contains out-of-date references to the old primary master. This causes package updates, like yum update, to fail because the repository is pointing to the old master. In general, the packaging configuration artifacts for PE infrastructure are now named puppet_enterprise across all platforms. The temporary artifacts created by the pe_repo module for agent installation are named pe_repo.

Permissions weren't set correctly on global Hiera file (2019.0.2)

On systems with a umask other than 000, the installation process didn't correctly set permissions on the global Hiera file at /etc/puppetlabs/puppet/global_hiera.yaml.

Package and service tasks run over SSH and WinRM relied on Puppet agent (2019.0.2)

The package and service tasks included with PE use implementations that do not require the Puppet agent be present on target nodes.

bolt-server clears temporary directory during cache purging (2019.0.2)

When pe-bolt-server purges the task cache directory, the temporary location where task downloads are staged is preserved. Previously the staging directory was deleted as part of the purge and this resulted in task download errors.

Deploying additional code or environments blocked Puppet Server startup (2019.0.1)

Using Code Manager to deploy additional code and environments increased the amount of time it took to start, restart, or reload the pe-puppetserver service, to the point that these operations could time out. The increase in startup time was caused by a diagnostic that checked every file in every environment and logged any modifications. This diagnostic has been removed.

Running r10k deploy on Bionic triggered a Ruby bug (2019.0.1)

Running either r10k 2.6 or 3.0 with the shellgit provider on Bionic could trigger a Ruby bug when running r10k deploy. This bug was more likely the more environments you deploy.

Slow response times when filtering nodes (2019.0.1)

Previously, when you filtered nodes based on fact values (a feature available from the Overview and Classification pages) in an environment with a large number of nodes, it would result in slow response times and, in some cases, cause timeout errors. Now when you filter nodes, a limited number of concurrent node queries is run, which improves performance.

Enumerated task parameter values not available in drop-down lists (2019.0.1)

When a single task parameter value that used the enumerated data type contained a hyphen (-), none of the set values appeared in drop-down lists. For example, these set values Enum[install, status-only, uninstall] would not appear because of the value status-only. This has been fixed.

Fedora 26 and 27 agent upgrade failed with module upgrade (2019.0.0)

Using the puppet-agent module to upgrade Fedora 26 and 27 agents failed.

Orchestrator errors for large task payloads (2019.0.0)

Previously, when running tasks that passed arguments larger than approximately 64KB, Orchestrator errors occurred. This was most likely to occur when using Bolt over the Puppet Communications Protocol to upload a file.

Back to top
The page rank or the 1 our of 5 rating a user has given the page.
The email address of the user submitting feedback.
The URL of the page being ranked/rated.