ModuleNotFoundError: No module named 'flask' on RHEL 8

[bschonec]

After days of attempting to solve the problem myself I'm forced to realize that I don't have the skills to debug Python or Apache.

On a brand new RHEL8 system with @minimal installed, I am unable to get Puppetboard installed properly due to an error about the flask module not being found.

Using a simple:

class { 'puppetboard':
  manage_git        => true,       
  manage_virtualenv => true,
}

Apache barks with:

[Tue Mar 29 08:57:24.729351 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516] mod_wsgi (pid=6074): Failed to exec Python script file '/srv/puppetboard/puppetboard/wsgi.py'.
[Tue Mar 29 08:57:24.729453 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516] mod_wsgi (pid=6074): Exception occurred processing WSGI script '/srv/puppetboard/puppetboard/wsgi.py'.
[Tue Mar 29 08:57:24.729764 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516] Traceback (most recent call last):
[Tue Mar 29 08:57:24.729823 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516]   File "/srv/puppetboard/puppetboard/wsgi.py", line 14, in <module>
[Tue Mar 29 08:57:24.729831 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516]     from puppetboard.app import app as application
[Tue Mar 29 08:57:24.729840 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516]   File "/srv/puppetboard/puppetboard/puppetboard/app.py", line 7, in <module>
[Tue Mar 29 08:57:24.729845 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516]     from flask import render_template, Response
[Tue Mar 29 08:57:24.729863 2022] [wsgi:error] [pid 6074] [remote 1.17.1.5:58516] ModuleNotFoundError: No module named 'flask'

Now, Puppetboard does seem to install properly but it's the Python module (I hope I'm describing it properly) that's giving me trouble.

Apache is running, but I get the following from curl/browser:

Internal Server Error
The server encountered an internal error or misconfiguration and was unable to complete your request.

Please contact the server administrator at [no address given] to inform them of the time this error occurred, and the actions you performed just before this error.

More information about this error may be available in the server error log.

A manual run of the WSGI yeilds:

[root@rhel8gold ~]# python3.8 /srv/puppetboard/puppetboard/wsgi.py
Traceback (most recent call last):
  File "/srv/puppetboard/puppetboard/wsgi.py", line 14, in <module>
    from puppetboard.app import app as application
  File "/srv/puppetboard/puppetboard/puppetboard/app.py", line 7, in <module>
    from flask import render_template, Response
ModuleNotFoundError: No module named 'flask'

It's been about four years since I've implemented a Puppet environment so I'm a bit rusty. I've tried the recommendations here:

voxpupuli/puppetboard#527

and here:

voxpupuli/puppetboard#582

but to no avail.

I've had even less success with RHEL7.

SELinux is off and the local firewall allows incoming TCP/80.

Here's my apache vhost config:

<VirtualHost *:80>
  ServerName rhel8gold-puppetboard

  ## Vhost docroot
  DocumentRoot "/srv/puppetboard/puppetboard"

  ## Directories, there should at least be a declaration for /srv/puppetboard/puppetboard

  <Directory "/srv/puppetboard/puppetboard">
    Options Indexes FollowSymLinks MultiViews
    AllowOverride None
    Require all granted
  </Directory>

  ## Logging
  ErrorLog "/var/log/httpd/rhel8gold-puppetboard_error.log"
  ServerSignature Off
  CustomLog "/var/log/httpd/rhel8gold-puppetboard_access.log" combined

  ## WSGI configuration
  WSGIDaemonProcess puppetboard group=puppetboard python-home=/srv/puppetboard/virtenv-puppetboard threads=5 user=puppetboard
  WSGIProcessGroup puppetboard
  WSGIScriptAlias / "/srv/puppetboard/puppetboard/wsgi.py"
</VirtualHost>

[bschonec]

Update:

If I set variable puppetboard::python_version: '3.6' then puppetboard does install properly and runs. The RedHat.8.yaml sets the python version to '3.8' and only when python is set to 3.6 does the virtual environment initialize properly.

[gdubicki]

Same as #333?

[bschonec]

Yep, nearly identical. When I changed the Python version to 3.6 everything worked. I'll rebuild my VM with python 3.8 and post the errors.

[jhunt-steds]

Hello everyone! I think I've chased down the issue here. Couple things to keep in mind:

• The RHEL 8 system version of Python is 3.6, so if you install the python3-whatever RPM, the Python whatever module gets installed in the Python 3.6 directories (e.g. /usr/lib64/python3.6/site-packages)
• As noted above, this module specifies Python 3.8 as its default desired version

So if you have a puppetboard::apache::vhost section on a RHEL server and take the defaults, this module:

1. Uses Python 3.8 in the virtual environment it sets up
2. Calls puppetlabs-apache's apache::mod::wsgi class with defaults (see e.g. puppetboard::apache::vhost lines 56-58)
3. apache::mod::wsgi installs the python3-mod_wsgi RPM, its default package for RHEL 8 (remember, this will be the Python 3.6 mod_wsgi package)
4. The puppet-python module gets invoked and tries to install python3 and python3-pip, which fails (but you can work around this by telling it to install whichever version of Python you really want, e.g.:

class { 'python':
  version => 'python38',
}

Note that this module requires you to specify the Python version including the decimal point between major and minor release, but the RHEL packages do not have a decimal.

The upshot of all this is that at first the modules don't even get installed in the virtual environment, and then if you fix the problem with the puppet-python package trying to install the wrong version, they are installed but don't work, because the mod_wsgi that's installed is looking for packages in the Python 3.6 paths (I got this error by making wsgi.py write out sys.path: [Tue Jun 14 14:27:32.182043 2022] [wsgi:error] [pid 3519] ['/srv/puppetboard/puppetboard', '/usr/lib64/python36.zip', '/usr/lib64/python3.6', '/usr/lib64/python3.6/lib-dynload']). If you manually uninstall the python3-mod_wsgi RPM and install (assuming Python 3.8) python38-mod_wsgi, Puppetboard starts working after an Apache restart (but of course Puppet hates that and so do I ;-) ). Thus, it seems to me the solution is for this module to be more specific about the package name on RHEL8 servers when calling apache::mod::wsgi, especially since thanks to DNF modules people might be using a variety of Python versions.

[jhunt-steds]

Hello again, I've found a workaround for the "wrong mod_wsgi" issue: use Hiera to declare some class defaults for apache::mod::wsgi. I added the following to my node's Hiera YAML file (continuing to use Python 3.8 as the example version):

---
apache::mod::wsgi::package_name: 'python38-mod_wsgi'
apache::mod::wsgi::mod_path: 'modules/mod_wsgi_python3.so'

...and now Puppet runs and sets up Puppetboard without issues. Note you do need both, if you try to specify just package_name it complains that it needs mod_path also. In case you prefer it, the full path to the module on RHEL is /usr/lib64/httpd/modules/mod_wsgi_python3.so. Hope this is helpful!