How to make ansible-provisioner not error out if a role is already installed

I am using the Ansible Packer provisioner to execute a playbook on one or more AWS instances brought up by Packer to build a custom Amazon Managed Image (AMI).

The instances are brought up in parallel but the issue is when I execute the Ansible provisioner it errors out in cases where it tries to reinstall roles that are already installed and asks me to use --ignore-errors.

amazon-ebs.build_ami: [WARNING]: - ansible_role_<REDACTED> was NOT installed successfully: the
amazon-ebs.build_ami: Starting galaxy role install process
amazon-ebs.build_ami: specified role ansible_role_<REDACTED> appears to already exist. Use
amazon-ebs.build_ami: --force to replace it.
amazon-ebs.build_ami: - extracting ansible_role_<REDACTED> to <REDACTED>/.ansible/roles/ansible_role_<REDACTED>
amazon-ebs.build_ami: ERROR! - you can use --ignore-errors to skip failed roles and finish processing the list.
==> amazon-ebs.build_ami: Provisioning step had errors: Running the cleanup provisioner, if present...
==> amazon-ebs.build_ami: Terminating the source AWS instance...

s there any way to stop the provisioner from trying to install the roles multiple times or send in --ignore-errors to the Ansible galaxy command used by the provisioner?

I have tried setting this up using the galaxy_command option but could not figure it out.

I also tried setting a custom roles_path and collections_path but the provisioner still installed the roles and collections to the default location specified by Ansible.

I would like to keep building the AMIs in parallel.

Here is a simplified version of build config:

build {
  dynamic "source" {
    for_each = local.images
    labels   = ["amazon-ebs.ami"]

    content {
      name       = "build_${source.value.value_one}_${source.value.value_two}_${source.value_three}"
      value_one   = source.value.value_one
      value_two = source.value.value_two
      value_three   = source.value.value_three
    }
 }

And here is a simplified version of ansible provisioner config:

provisioner "ansible" {
    playbook_file  = "playbook.yml"
    galaxy_file    = "requirements.yml"
    use_proxy      = false

    ansible_env_vars = [
      "ANSIBLE_FORCE_COLOR=1",                 # Force colored output
      "ANSIBLE_PYTHON_INTERPRETER=auto_silent" # Silence warning about Python discovery
    ]

    extra_arguments = [
      "--extra-vars", "\"ansible_python_interpreter=/usr/bin/env python3\"", # Find Python 3 on PATH
    ]
  }

I hope someone could give me a way to either use --ignore-errors which is not ideal or some other way to get around this issue.

This is somewhat of a known conflict between ansible-galaxy and packer-plugin-ansible due to issues related to caching. I usually solve this with the galaxy_force_install parameter:

provisioner "ansible" {
  playbook_file        = "playbook.yml"
  galaxy_file          = "requirements.yml"
  galaxy_force_install = true
  use_proxy            = false

  ansible_env_vars = [
    "ANSIBLE_FORCE_COLOR=1",                 # Force colored output
    "ANSIBLE_PYTHON_INTERPRETER=auto_silent" # Silence warning about Python discovery
  ]

  extra_arguments = [
    "--extra-vars", "\"ansible_python_interpreter=/usr/bin/env python3\"", # Find Python 3 on PATH
  ]
}

There is a chance this may still cause issues for you with role and collection dependencies, and in that situation I would recommend going the full force with galaxy_force_with_deps:

provisioner "ansible" {
  playbook_file          = "playbook.yml"
  galaxy_file            = "requirements.yml"
  galaxy_force_with_deps = true
  use_proxy              = false

  ansible_env_vars = [
    "ANSIBLE_FORCE_COLOR=1",                 # Force colored output
    "ANSIBLE_PYTHON_INTERPRETER=auto_silent" # Silence warning about Python discovery
  ]

  extra_arguments = [
    "--extra-vars", "\"ansible_python_interpreter=/usr/bin/env python3\"", # Find Python 3 on PATH
  ]
}

You can view more information about these optional parameters in the documentation.