ansible_local's pip installer has a hardcoded curl get_pip.py piped to python

[chadswen]

When the ansible_local provisioner configures ansible.install_mode = pip this command is invoked:

vagrant/plugins/provisioners/ansible/cap/guest/pip/pip.rb

Line 23 in cb61497

machine.communicate.execute "curl https://bootstrap.pypa.io/get-pip.py | sudo python"

Piping curl into python with a static command prevents me from setting get_pip.py args. In my case I need to set a proxy, but I don't want to set a proxy for the entire provisioner as that causes issues when running certain playbooks. As far as I can tell there's no way to set a proxy for only the ansible install phase of the provisioner.

As a workaround, I use a shell provisioner to install pip manually with my custom get_pip.py args before ansible_local runs, but I still have to wait for ansible_local's static curl | python command to timeout on every install which is fairly annoying. I would use the package manger ansible install_mode, but my ansible playbooks require specific ansible versions and you can only target an ansible version with the pip install_mode.

Vagrant version

Vagrant 2.0.2

[gildegoma]

@chadswen Thank you for reporting this use case. The (Ansible) provisioners are not aiming at supporting any use case, and for "advanced/complex" situations we try to find the adequate balance (code complexity, ROI, Vagrant flexibility for customization, etc.).

In my case I need to set a proxy, but I don't want to set a proxy for the entire provisioner as that causes issues when running certain playbooks.

Running behind a proxy is typically subject to variance and somehow hard to fully cover by relying on standard configuration capabilities (for instance the Docker provisioner also installs with hardcoded curl commands).

So, you did the right thing with implementing your own ansible setup via a shell provisioner:

As a workaround, I use a shell provisioner to install pip manually with my custom get_pip.py args before ansible_local runs, but I still have to wait for ansible_local's static curl | python command to timeout on every install which is fairly annoying.

Now, to avoid this useless and annoying wait, you can extend your shell provisioner to also install ansible, and then configure the ansible_local provisioner to not install ansible by setting the install option to false. The provisioner will then only verify that the expected version is available.

Please let me know if this solution works for you and if you agree to close this issue (i.e. no feature request). Thanks!

[chadswen]

@gildegoma I'd hate to stop using the ansible installer. It installs the specific version of ansible as required by my playbook and that feature saves me from rewriting a fair amount of code with a shell provisioner.

My only problem is that there is no way to set args when ansible_local installs pip before it installs ansible. I can set args just fine when it installs ansible using ansible.pip_args, but the curl | python command I linked from the OP is hardcoded to always pull from the internet without any configuration options for how that is done.

[gildegoma]

The issue is quite clear to me, and I do appreciate that the existing features are used ;-)

that feature saves me from rewriting a fair amount of code with a shell provisioner

I guess it shouldn't be that large (e.g. pip install ...), but I ear your "frustration" to be so close from the goal without having a hook to catch.

@chadswen what would be your proposal to resolve this?

I would propose to introduce a new pip_install_command option, with
"curl https://bootstrap.pypa.io/get-pip.py | sudo python" as default.
What do you think? Interested to work on it?

[chadswen]

@gildegoma That solution would work quite well. I can't volunteer to work on it right now, but perhaps in the future if it isn't picked up.

Thanks!

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.