Troubleshooting Code Manager

Code Manager requires coordination between multiple components, including r10k and the file sync service. If you have issues with Code Manager, check that these components are functioning.

Code Manager logs

Code Manager logs to the Puppet Server log. By default, this log is in /var/log/puppetlabs/puppetserver/puppetserver.log. For more information about working with the logs, see the Puppet Server logs documentation.

Check Code Manager status

Check the status of Code Manager and file sync if your deployments are not working as expected, or if you need to verify that Code Manager is enabled before running a dependent command.

The command puppet-code status verifies that Code Manager and file sync are responding. The command returns the same information as the Code Manager status endpoint. By default, the command returns details at the info level; critical and debug aren’t supported.

The following table shows errors that might appear in the puppet-code status output.
Error Cause
Code Manager couldn’t connect to the server pe-puppetserver process isn’t running
Code Manager reports invalid configuration Invalid configuration at /etc/puppetlabs/puppetserver/conf.d/code-manager.conf
File sync storage service reports unknown status Status callback timed out

Test the connection to the control repository

The control repository controls the creation of environments, and ensures that the correct versions of all the necessary modules are installed. The primary server must be able to access and clone the control repo as the pe-puppet user.

To make sure that Code Manager can connect to the control repo, run:

puppet-code deploy --dry-run

If the connection is set up correctly, this command returns a list of all environments in the control repo or repos. If the command completes successfully, the SSH key has the correct permissions, the Git URL for the repository is correct, and the pe-puppet user can perform the operations involved.

If the connection is not working as expected, Code Manager reports an Unable to determine current branches for Git source error.

The unsuccessful command also returns a path on the primary server that you can use for debugging the SSH key and Git URL.

Check the Puppetfile for errors

Check the Puppetfile for syntax errors and verify that every module in the Puppetfile can be installed from the listed source. To do this, you need a copy of the Puppetfile in a temporary directory.

Create a temporary directory /var/tmp/test-puppetfile on the primary server for testing purposes, and place a copy of the Puppetfile into the temporary directory.

You can then check the syntax and listed sources in your Puppetfile.

Check Puppetfile syntax

To check the Puppetfile syntax, run r10k puppetfile check from within the temporary directory.

If you have Puppetfile syntax errors, correct the syntax and test again. When the syntax is correct, the command prints "Syntax OK".

Check the sources listed in the Puppetfile

To test the configuration of all sources listed in your Puppetfile, perform a test installation. This test installs the modules listed in your Puppetfile into a modules directory in the temporary directory.

In the temporary directory, run the following command:

sudo -H -u pe-puppet bash -c \
  '/opt/puppetlabs/puppet/bin/r10k puppetfile install'

This installs all modules listed in your Puppetfile, verifying that you can access all listed sources. Take note of all errors that occur. Issues with individual modules can cause issues for the entire environment. Errors with individual modules (such as Git URL syntax or version issues) show up as general errors for that module.

If you have modules from private Git repositories requiring an SSH key to clone the module, check that you are using the SSH Git URL and not the HTTPS Git URL.

After you've fixed errors, test again and fix any further errors, until all errors are fixed.

Run a deployment test

Manually run a full r10k deployment to check not only the Puppetfile syntax and listed host access, but also whether the deployment works through r10k.

This command attempts a full r10k deployment based on the r10k.yaml file that Code Manager uses. This test writes to the code staging directory only. This does not trigger a file sync and is to be used only for ad-hoc testing.

Run this deployment test with the following command:

  sudo -H -u pe-puppet bash -c \
    '/opt/puppetlabs/puppet/bin/r10k deploy environment -c /opt/puppetlabs/server/data/code-manager/r10k.yaml -p -v debug'

If this command completes successfully, the /etc/puppetlabs/code-staging directory is populated with directory-based environments and all of the necessary modules for every environment.

If the command fails, the error is likely in the Code Manager settings specific to r10k. The error messages indicate which settings are failing.

Monitor logs for webhook deployment trigger issues

Issues that occur when a Code Manager deployment is triggered by the webhook can be tricky to isolate. Monitor the logs for the deployment trigger to find the issue.

Deployments triggered by a webhook in Stash/Bitbucket, GitLab, or GitHub are governed by authentication and hit each service's /v1/webhook endpoint.

If you are using a GitLab version older than 8.5.0, Code Manager webhook authentication does not work because of the length of the authentication token. To use the webhook with GitLab, either disable authentication or update GitLab.

Note: If you disable webhook authentication, it is disabled only for the /v1/webhook endpoint. Other endpoints (such as /v1/deploys) are still controlled by authentication. There is no way to disable authentication on any other Code Manager endpoint.

To troubleshoot webhook issues, follow the Code Manager webhook instructions while monitoring the Puppet Server log for successes and errors. To do this, open a terminal window and run:

  tail -f /var/log/puppetlabs/puppetserver/puppetserver.log

Watch the log closely for errors and information messages when you trigger the deployment. The puppetserver.log file is the only location these errors appear.

If you cannot determine the problem with your webhook in this step, manually deploy to the /v1/deploys endpoint, as described in the next section.

Monitor logs for /v1/deploys deployment trigger issues

Issues that occur when a Code Manager deployment is triggered by the /v1/deploys endpoint can be tricky to isolate. Monitor the logs for the deployment trigger to find the issue.

Before you trigger a deployment to the /v1/deploys endpoint, generate an authentication token. Then deploy one or more environments with the following command:

type_header='Content-Type: application/json'
auth_header="X-Authentication: $(puppet-access show)"
uri="https://$(puppet config print server):8170/code-manager/v1/deploys"
data='{"environments": ["production"], "wait": true}'

curl --insecure --header "$type_header" --header "$auth_header" --request POST "$uri" --data "$data"

See Usage notes for curl examples for information about forming curl commands.

This request waits for the deployment and sync to compilers to complete ("wait": true) and so returns errors from the deployment.

Alternatively, you can deploy all environments with the following curl command:

type_header='Content-Type: application/json'
auth_header="X-Authentication: $(puppet-access show)"
uri="https://$(puppet config print server):8170/code-manager/v1/deploys"
data='{"deploy-all": true, "wait": true}'

curl --insecure --header "$type_header" --header "$auth_header" --request POST "$uri" --data "$data"

Monitor the console-services.log file for any errors that arise from this curl command.

tail -f /var/log/puppetlabs/console-services/console-services.log

Code deployments time out

If your environments are heavily loaded, code deployments can take a long time, and the system can time out before deployment is complete.

If your deployments are timing out too soon, increase your timeouts_deploy key. You might also need to increase:

  • timeouts_shutdown
  • timeouts_sync
  • timeouts_wait

File sync stops when Code Manager tries to deploy code

Code Manager code deployment involves accessing many small files. If you store your Puppet code on network-attached storage, slow network or back-end hardware can result in poor deployment performance.

Tune the network for many small files, or store Puppet code on local or direct-attached storage.

Classes are missing after deployment

After a successful code deployment, a class you added isn't showing in the console.

If your code deployment works, but a class you added isn't in the console:

  1. Check on disk to verify that the environment folder exists.
  2. Check your node group in the Edit node group metadata box to make sure it's using the correct environment.
  3. Refresh classes.
  4. Refresh your browser.