Agent installation fails on aws amzn2

[msiebeneicher]

Hi,

we are using the datadog.datadog role from ansible galaxy now for a couple of time.
Since the last agent version 7.32.4 (how it looks like) our packer build is failing with the following error message:

TASK [datadog.datadog : Fail early if Python 3 is used on CentOS / RHEL < 8] ***
    amz-ami: fatal: [default]: FAILED! => {"changed": false, "msg": "The installation of the Agent on CentOS / RHEL versions < 8 requires the 'yum' module, which is not compatible with Python 3. To run this role, use a Python 2 interpreter on hosts running CentOS / RHEL < 8."}

This error happen on the latest aws ami amzn2-ami-hvm-2.0.20211201.0-x86_64-ebs (ami-01826408c6676f97e) and also on the previous version (ami-0d455065dfd31a3cd) which we already successfully used in the past.

As a workaround we tried successfully to install the agent by the official shell installer on the same ec2 instance:
DD_AGENT_MAJOR_VERSION=7 DD_API_KEY=<KEy> DD_SITE="datadoghq.eu" bash -c "$(curl -L https://s3.amazonaws.com/dd-agent/scripts/install_script.sh)"

I also tried the workaround described in your docs under https://docs.datadoghq.com/agent/basic_agent_usage/ansible/ which also was not working in our case because thednf and the python3-dnf packages are not available in the default aws repositories. If I use the optiondatadog_ignore_old_centos_python3_error: true i get the following error message:

TASK [datadog.datadog : Install latest datadog-agent package (dnf)] ************
    amz-ami: fatal: [default]: FAILED! => {"changed": false, "msg": "Could not import the dnf python module using /usr/bin/python3.7 (3.7.10 (default, Jun  3 2021, 00:02:01) [GCC 7.3.1 20180712 (Red Hat 7.3.1-13)]). Please install `python3-dnf` or `python2-dnf` package or ensure you have specified the correct ansible_python_interpreter. (attempted ['/usr/libexec/platform-python', '/usr/bin/python3', '/usr/bin/python2', '/usr/bin/python'])", "results": []}

Any ideas or suggestions to handle this issue or why this issue is popping up now?

Here my used version:

• ansible [core 2.12.1]
• packer 1.7.8

datadog.datadog role version:

• 4.12.0
• 4.11.0

aws amis:

• "Name": "amzn2-ami-hvm-2.0.20211201.0-x86_64-ebs",
  "ImageId": "ami-01826408c6676f97e",
  "CreationDate": "2021-12-01T19:55:07.000Z"

• "Name": "amzn2-ami-hvm-2.0.20211005.0-x86_64-ebs",
  "ImageId": "ami-0d455065dfd31a3cd",
  "CreationDate": "2021-10-12T14:12:46.000Z"

[moleskin-smile]

I have the same problem on Amazon Linux 2.

So the problematic change is this one: #301.

I think it is based on assumptions that are not always true. It wants to use dnf intead of yum if ansible_facts.python.version.major >= 3 – I don't think this is the right thing to do. ansible_facts.python.version.major is just a Python version detected by Ansible. It doesn't mean it's a default Python installation, nor allows to conclude about package manager being used. We run Ansible with Python 3 on Amazon Linux 2 without problems for years.

I think the right fact to use is ansible_pkg_mgr.

Side note: the yum Ansible builtin supports dnf, see:

This module supports yum (as it always has), this is known as yum3/YUM3/yum-deprecated by upstream yum developers. As of Ansible 2.7+, this module also supports YUM4, which is the "new yum" and it has an dnf backend.
By default, this module will select the backend based on the ansible_pkg_mgr fact.

So maybe even no changes are required. And the original issue (#298), that reported this, mentions documentation entry about yum module not being supported on Python 3. But it is – for years. Maybe not via native bindings, but it works. And regarding dnf support – AFAIK there's compatibility layer that allows using yum command on a dnf system.

[moleskin-smile]

@msiebeneicher: could you please test your setup with changes from #412? Thanks!