puppet infrastructure run commands can fail if the agent is run with cron

puppet infrastructure run commands, such as those used for certain installation, upgrade, and certificate management tasks, can fail if the Puppet agent is run with cron. The failure occurs if the command conflicts with a Puppet run. As a workaround, you can either disable the agent running out of cron on the nodes the plan affects, or be careful that cron doesn't execute an agent run while the command is running.

Installer can fail due to SSL errors with AmazonAWS

In some cases when attempting to install PE, some master platforms have received SSL errors when attempting to connect to s3.amazonaws.com, and thus have been unable retrieve puppet-agent packages needed for installation. In most cases, you should be able to properly install after updating the CA cert bundle on the master platform. To update the bundle, run the following commands:

rm /etc/ssl/certs/ca-bundle.crt
yum reinstall ca-certificates

After updating the CA bundle, run the PE installer again.

Errors when using puppet code and puppet db commands on FIPS-enabled hardware

/opt/puppetlabs/puppet/bin/gem install puppetdb_cli --bindir /opt/puppetlabs/bin/

FIPS installations do not support Razor

The FIPS platform does not support the Razor component in PE.

puppet infrastructure recover_configuration misreports success if specified environment doesn't exist

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

Restoring the pe-rbac database fails with the puppet-backup restore command

When restoring the pe-rbac database, the restore process exits with errors about a duplicate operator family, citext_ops.

puppet-backup fails if gems are missing from the master's agent environment

The puppet-backup create command might fail if any gem installed on the Puppet Server isn't present on the agent environment on the master. If the gem is either absent or of a different version on the master's agent environment, you get the error "command puppet infrastructure recover_configuration failed".

To fix this, you must install any missing or incorrectly versioned gems on the master's agent environment. To find which gems are causing the error, check the backup logs for any gem incompatibility issues with the error message. PE creates backup logs as a report.txt whenever you run a puppet-backup command.

To see what gems and their versions you have installed on your Puppet Server, run the command puppetserver gem list . To see what gems are installed in the agent environment on your master, run /opt/puppetlabs/puppet/bin/gem list.

Console is inaccessible on macOS Catalina using default certificates

puppet_enterprise::profile::console::browser_ssl_cert: /opt/puppetlabs/server/data/console-services/certs/public-console.cert.pem
puppet_enterprise::profile::console::browser_ssl_private_key: /opt/puppetlabs/server/data/console-services/certs/public-console.private_key.pem

Mismatch between classifier classification and matching nodes for regexp rules

PuppetDB’s regular expression matching has surprising behaviors for structured fact value comparisons. For example, the structured fact os is a rule that matches ["~", "os", ":"]. PuppetDB would unintentionally match every node that has the os structured fact because the regular expression is applied to the JSON encoded version of the fact value.

The classifier does not use PuppetDB for determining classification and regular expressions in the classifier rules syntax only support direct value comparisons for string types.

This has caused issues in the console where the node list and counts for the "matching nodes" display sometimes indicated that nodes were matching even though the classifier would not consider them matching.

Orchestrator fails when rerunning tasks on agentless nodes

When you rerun tasks from the Job details page, target nodes that do not have the Puppet agent installed are miscategorized as PuppetDB nodes. This causes the orchestrator to fail on those nodes.

Target name localhost unavailable for plans in PE

The special target name localhost is not available in plans for PE since the definition is ambiguous.

Some Bolt stdlib functions fail in PE

For more information, see Plan caveats and limitations

No metadata in puppet plan show <PLAN NAME> function

There is a known issue where the puppet plan show <PLAN NAME> function does not return plan metadata. To see plan metadata, view the plan source code.

The function puppet plan show displays invalid plans when Code Manager is disabled

If Code Manager is disabled, the puppet plan show function and the PE console will still list plans that are contained in a control repo, but those plans cannot actually be run. Enable Code Manager to run plans.

Regenerating agent certificates fails with autosign enabled

The puppet infrastructure run regenerate_agent_certificate command includes a step for signing the node's certificate. With autosign enabled, an unsigned certificate can't be found, and the command errors out. As a workaround, temporarily disable autosign before running puppet infrastructure run regenerate_agent_certificate.

Default SSH URL with TFS fails with rugged error

Using the default SSH URL with Microsoft Team Foundation Server (TFS) with the rugged provider causes an error of "unable to determine current branches for Git source." This is because the rugged provider expects an @ symbol in the URL format.

To work around this error, replace ssh:// in the default URL with git@

ssh://tfs.puppet.com:22/tfs/DefaultCollection/Puppet/_git/control-repo

git@tfs.puppet.com:22/tfs/DefaultCollection/Puppet/_git/control-repo

GitHub security updates might cause errors with shellgit

GitHub has disabled TLSv1, TLSv1.1 and some SSH cipher suites, which can cause automation using older crypto libraries to start failing. If you are using Code Manager or r10k with the shellgit provider enabled, you might see negotiation errors on some platforms when fetching modules from the Forge. To resolve these errors, switch your configuration to use the rugged provider, or fix shellgit by updating your OS package.

Timeouts when using --wait with large deployments or geographically dispersed compilers

Because the --wait flag now deploys code to all compilers before returning results, some deployments with a large node count or compilers spread across a large geographic area might experience a timeout. Work around this issue by adjusting the timeouts_sync parameter.

r10k with the Rugged provider can develop a bloated cache

If you use the Rugged provider for r10k, repository pruning is not supported. As a result, if you use many short-lived branches, over time the local r10k cache can become bloated and take up significant disk space.

If you encounter this issue, run git-gc periodically on any cached repo that is using a large amount of disk space in the cachedir. Alternately, use the shellgit provider, which automatically garbage collects the repos according to the normal Git CLI rules.

Code Manager and r10k do not identify the default branch for module repositories

When you use Code Manager or r10k to deploy modules from a Git source, the default branch of the source repository is always assumed to be master. If the module repository uses a default branch that is not master, an error occurs. To work around this issue, specify the default branch with the ref: key in your Puppetfile.

After an error during the initial run of file sync, Puppet Server won't start

The first time you run Code Manager and file sync on a master, an error can occur that prevents Puppet Server from starting. To work around this issue:

1. Stop the pe-puppetserver service.
2. Locate the data-dir variable in /etc/puppetlabs/puppetserver/conf.d/file-sync.conf.
3. Remove the directory.
4. Start the pe-puppetserver service.

Repeat these steps on each master exhibiting the same symptoms, including any compilers.

Puppet Server crashes if file sync can't write to the live code directory

If the live code directory contains content that file sync didn’t expect to find there (for example, someone has made changes directly to the live code directory), Puppet Server crashes.

The following error appears in puppetserver.log:

2016-05-05 11:57:06,042 ERROR [clojure-agent-send-off-pool-0] [p.e.s.f.file-sync-client-core] Fatal error during file sync, requesting shutdown.
org.eclipse.jgit.api.errors.JGitInternalException: Could not delete file /etc/puppetlabs/code/environments/development
        at org.eclipse.jgit.api.CleanCommand.call(CleanCommand.java:138) ~[puppet-server-release.jar:na]

To recover from this error:

1. Delete the environments in code dir: find /etc/puppetlabs/code -mindepth 1 -delete.
2. Start the pe-puppetserver service: puppet resource service pe-puppetserver ensure=running
3. Trigger a Code Manager run by your usual method.

Code manager cannot deploy forge modules with a proxy (2019.2.0 and 2019.2.1)

The commands puppet code deploy and r10k fail when behind a proxy. The commands do not use the configured proxy settings and using them can result in issues downloading modules from the Puppet forge. This is due to an issue in a dependency gem.

/opt/puppetlabs/puppet/bin/gem install faraday -v 0.12.2
/opt/puppetlabs/puppet/bin/gem uninstall faraday -v 0.13.1
puppet resource service pe-puppetserver ensure=stopped
puppet resource service pe-puppetserver ensure=running

ASCII limitations

Ruby can corrupt the path fact and environment variable on Windows

There is a bug in Ruby that can corrupt the environment variable names and values. This bug causes corruption for only some codepages. This bug might be triggered when environment names or values contain UTF-8 characters that can't be translated to the current codepage.

The same bug can cause the path fact to be cached in a corrupt state.