Terraform Enterprise (Replicated): tfe-atlas container restarts continuously with command not found: puma – HashiCorp Help Center

Problem

The tfe-atlas container in a Terraform Enterprise instance continuously restarts.

Prerequisites

An installation of Terraform Enterprise (Replicated deployment) with non-consolidated services.
Docker Engine 1.13.1 from the Extra Packages for Enterprise Linux (EPEL) repository using a modified libseccomp profile.

Cause

A misconfiguration of the libseccomp settings for Docker can cause this issue. This misconfiguration prevents the puma command from being found within the tfe-atlas container, leading to the following error and a restart loop.

bundler: command not found: puma
Install missing gem executables with `bundle install`

Solutions

Solution 1: Modify the Docker seccomp Profile

To resolve the container restart loop, you must adjust the default action in the Docker seccomp.json file.

Execute the following command to modify the seccomp.json file, changing the defaultAction from SCMP_ACT_ERRNO to SCMP_ACT_TRACE.
```
$ sudo sed -i 's/"defaultAction":\s*"SCMP_ACT_ERRNO"/"defaultAction": "SCMP_ACT_TRACE"/1' /etc/docker/seccomp.json
```
Restart the Docker service and Terraform Enterprise for the changes to take effect.

Outcome

After applying the configuration change and restarting the services, the tfe-atlas container should start successfully and no longer enter a restart loop.

Additional Information

For more details on Docker requirements for Terraform Enterprise, refer to the official documentation on Docker Engine Using a Modified Libseccomp Profile.