Mitogen does not support Ansible 2.8 Python interpreter detection

[ghost]

Running Debian 9 locally with Ansible 2.8.1 with Python 3. Controlling a Ubuntu 18.04 server.

I followed the instructions here: https://networkgenomics.com/ansible/

Of course it didn't work for 2.8.X, so I switched it out for master. Now I'm getting this:

TASK [Gathering Facts] ************************************************************************************************************************
fatal: [8.9.11.101]: UNREACHABLE! => {"changed": false, "msg": "EOF on stream; last 300 bytes received: 'bash: /usr/bin/python: No such file or directory\\n'", "unreachable": true}

Oddly enough, /usr/bin/python does exist but it's a symlink to python2.7.

I think it should be trying /usr/bin/python3. Ansible works just fine without mitogen enabled.

I think there's something about a ansible_python_interpreter that I can set, but seems like this should work by default.

What can I try/test? I'm new to both Ansible and Mitogen, so not very familiar with either yet.

Thank you!

[dw]

Hi there,

This is probably something to do with the new "Python interpreter discovery" feature in Ansible 2.8. Mitogen currently lacks support for it.

As a workaround, you can specify ansible_python_interpreter=/usr/bin/python3 for just that machine, or for its group, or simply on the command line using ansible -e ansible_python_interpreter=...

Sorry for the hassle! Interpreter discovery is simultaneously a simple feature and fairly painful, because to solve it requires implementing raw: module support, which Mitogen has always lacked. Coming soon!

[ghost]

Thank you, that works!

Actually ended up setting that yesterday because Ansible was doing Python 2 on one host and Python 3 on another.

Do you want me to close this out or leave it open for now?

[dw]

All behavioural differences are bugs, so it stays open :) Thanks for reporting!

[ghost]

No problem, sounds good!

[rei-android]

Any news on this?

[s1113950]

@dw I got a different error with Ansible 2.8.5 and Mitogen 0.2.8:

fatal: [machine]: UNREACHABLE! => {"changed": false, "msg": "EOF on stream; last 100 lines received:\nusername@{machineIP}'s password: ", "unreachable": true}

task:

- name: pull down docker binaries
  get_url:
    url: https://download.docker.com/linux/static/stable/x86_64/docker-{{ docker_version }}.tgz
    dest: /usr/local/share
    mode: 0644
  vars:
    ansible_python_interpreter: source /opt/rh/rh-python36/enable && python

In this case a machine's default system python was super old so we needed to use a user-installed python instead to avoid SNI issues. This works just fine without mitogen.

[s1113950]

I believe I discovered my issue. It's due to how the ssh command is set up when ansible_python_interpreter is more than just a python path:

[mux  96418] 17:02:52.927805 D mitogen.parent: command line for Connection(None): ssh -o "LogLevel ERROR" -l {login_name} -o "Compression yes" -o "ServerAliveInterval 30" -o "ServerAliveCountMax 10" -o "StrictHostKeyChecking no" -o "UserKnownHostsFile /dev/null" -o "GlobalKnownHostsFile /dev/null" -o ForwardAgent=yes {IP_ADDR} source /opt/rh/rh-python36/enable '&&' python -c "'import codecs,os,sys;_=codecs.decode;exec(_(_(\"eNqFkUFPwzAMhc/...

This is different than what ansible does in this situation:

8
SSH: EXEC sshpass -d41 ssh -vvv -o ForwardAgent=yes -o StrictHostKeyChecking=no -o 'User="{username}"' -o ConnectTimeout={IP_ADDR} '/bin/sh -c '"'"'sudo -H -S  -p "[sudo via ansible, key=SOMEKEY] password:" -u root /bin/sh -c '"'"'"'"'"'"'"'"'echo BECOME-SUCCESS-{SOMEKEY} ; source /opt/rh/rh-python36/enable && python'"'"'"'"'"'"'"'"' && sleep 0'"'"''

I'll try and submit a PR to fix

[s1113950]

Will be working on the fix here: #658. Might be going down the rabbit hole a bit but I'm determined 😅

[s1113950]

My fix is ready

[ssbarnea]

Really sad to see that there is no progress on this bug, which is a deal breaker.

[snoord]

Facing the same issue. I have managed to work around this issue by setting the following for my CentOS 7 and CentOS 8 group:

ansible_python_interpreter: /usr/libexec/platform-python

This has allowed me to continue working with both sets of VMs at a time, as when I had previously set the python interpreter to /usr/bin/python3, I lost the ability to use the yum module with the CentOS 7 hosts.

[s1113950]

I've made a PR here: #658 can those affected please try my patch and see if it fixes things for you?