ansible-local: setting PATH and using galaxy_command

[Ed Jackson]

Hi all,

I'm a bit of a Packer n00b, so I hope this isn't a silly question. Skip to the bottom for the tl;dr, but here's the situation:

I'm using the ansible-local provisioner on an image where ansible was installed in the ssh user's account via

pip install --user ansible~=2.9

This places all the ansible executables in ~/.local/bin.

The user's $PATH is modified appropriately in the bash profile, but since Packer's ssh communicator doesn't run the remote command in a login shell, this has no effect. If I update the command to be

PATH=$PATH:~/.local/bin ANSIBLE_REMOTE_TEMP=/opt/packer-provisioner-ansible-local PYTHONUNBUFFERED=1 ansible-playbook

it seems to work fine (of course I could just as easily specify the full command path).

The problem comes when I try to run ansible-galaxy. I get a command not found error, so I attempt a similar fix by setting galaxy_command:

PATH=$PATH:~/.local/bin ansible-galaxy

This is documented here, but I get the following error:

Errors validating build 'mybuild'. 1 error occurred:
    * unknown configuration key: "galaxy_command"

I am running packer version 1.4.5; it looks like the galaxy parameters were added in 1.4.3. The target OS is Amazon Linux 2 (very much RHEL compatible).

1) Why can't I use galaxy_command?

2) Is there a better way of handling the non-standard location for ansible binaries?

Thanks!

Ed

[Ed Jackson]

A co-worker found this for me:

https://github.com/hashicorp/packer/pull/5105

Apparently the key is "galaxycommand" even though it's documented as "galaxy_command".

My second question still stands if anyone has input.