`No inventory was parsed` for aap/awx

[jiholland]

In AAP/AWX, all of my check_point.gaia-related playbooks fail at the very first task with the error message:

Task failed: No inventory was parsed, please check your configuration and options.

This issue started around October 2025. It does not seem to matter what I change in the execution environment — I have tried numerous variations without success. Gaia is the only device type affected. Playbooks using check_point.mgmt, cisco.ios, community.zabbix, and cisco.nxos all work correctly with the same inventory source and execution environment. It also appears that others are experiencing the same issue: https://forum.ansible.com/t/no-inventory-was-parsed-in-awx-but-works-with-ansible-navigator/42648

I'm using Python 3.12.12, ansible-core 2.20.2, ansible.netcommon 8.4.0, ansible.utils 6.0.1 and check_point.gaia 7.0.0 - both in CLI and the execution environment for aap/awx. The gaia collection is working without any problems in CLI and also if I pull the execution environment image, run it and execute gaia related playbooks from within it. It only fails with the No inventory parsed message if I launch jobs from AAP/AWX.

Any help or guidance would be greatly appreciated.

[jiholland]

This has to do with https://github.com/CheckPointSW/CheckPointAnsibleGAIACollection/blob/master/plugins/httpapi/checkpoint.py

class HttpApi(HttpApiBase):
    def __init__(self, connection):
        super(HttpApi, self).__init__(connection)
        self.connection = connection
        self.mgmt_proxy_enabled = False

        loader = DataLoader()
        # Initialize InventoryManager
        inventory = InventoryManager(loader=loader, sources=['/etc/ansible/hosts'])
        # Get host
        host = inventory.get_host('mgmt_proxy')
        # Get variable
        try:
            proxy_enabled = host.vars['enabled']
            if proxy_enabled is True:
                self.mgmt_proxy_enabled = True
        except Exception as e:
            pass

The plugin always tries to load /etc/ansible/hosts without checking if it exists. If the file is missing, InventoryManager returns an empty inventory, which causes Ansible to report No inventory was parsed. Creating an empty /etc/ansible/hosts file in the execution environment "solved" the issue, but it is probably better to move the inventory logic out of the plugin. Perhaps something like this:

  mgmt_proxy_enabled:
    type: bool
    default: false
    description:
      - Enable management proxy
    vars:
      - name: ansible_checkpoint_mgmt_proxy_enabled

class HttpApi(HttpApiBase):
    def __init__(self, connection):
        super().__init__(connection)
        self.connection = connection
        self.mgmt_proxy_enabled = self.get_option("mgmt_proxy_enabled")

[ktbyers]

@jiholland I ran into this same issue on Gaia Version 7.0 and saw the same problem i.e. the attempt to load inventory from a non-existent /etc/ansible/hosts. This basically broke the AWX execution environment.

This same issue also causes the following (somewhat strange) warning outside of AWX:

$ ansible-playbook show_version.yml 

PLAY [Basic Gaia Playbook] ******************************************************************************

TASK [Show Version] *************************************************************************************
[WARNING]: No inventory was parsed, only implicit localhost is available
ok: [pod2-gaia]

Outside of AWX, it only gave you a warning and didn't completely stop the playbook from executing (so I largely ignored the warning).

[ktbyers]

This entire inventory loading should go away (line 56) i.e. the inventory should be passed to the connection plugin and not-reloaded.

 48 class HttpApi(HttpApiBase):
 49     def __init__(self, connection):
 50         super(HttpApi, self).__init__(connection)
 51         self.connection = connection
 52         self.mgmt_proxy_enabled = False
 53 
 54         loader = DataLoader()
 55         # Initialize InventoryManager      ### The plugin should NOT be trying to re-load inventory 
 56         inventory = InventoryManager(loader=loader, sources=['/etc/ansible/hosts'])
 57         # Get host
 58         host = inventory.get_host('mgmt_proxy')
 59         # Get variable
 60         try:
 61             proxy_enabled = host.vars['enabled']
 62             if proxy_enabled is True:
 63                 self.mgmt_proxy_enabled = True
 64         except Exception as e:
 65             pass

Right now the code looks for a host named mgmt_proxy and then if that exists you look for a variable named enabled.

It would (probably) be more consistent with Ansible patterns to not have this triggered based on a specific host existing and instead to have these as Ansible host/group variables i.e. something like:

all:
  vars:
    # Set this for "all" or for a group or for a specific host
    use_mgmt_proxy: true 

This would imply that the actual "ansible_host" for each Gateway was the Mgmt Proxy server.

I haven't worked through the entire workflow so definitely could be missing something (but the above pattern is more consistent with what Ansible expects i.e. this mgmt_proxy should be a host/group variable bound to each inventory member).