SUMMARY

There seems to be a race condition when provisioning a new device and waiting for the provisioning to complete (either via setting state: active or via wait_for_public_IPv: 4/6). Specifically, if the device disappears from the active devices list while waiting for either of those conditions, the equinix.metal.device module will success with an empty device list instead of failing. This can happen, for example, when attempting to provision a new host with the same hostname as a host in the process of being de-provisioned.

This seems to be due to the fact that the wait_for_devices_active and wait_for_public_IPv functions attempt to wait for their conditions to be met by polling the results refresh_devices_list and checking that all devices returned match their requirements, but refresh_devices_list only returns devices visible from the API without checking to see if that includes all the device IDs passed as a parameter.

ISSUE TYPE

• Bug Report

COMPONENT NAME

equinix.metal.device

STEPS TO REPRODUCE

1. Provision a server with the following play:

- name: Provision an Equinix server.
  register: output
  equinix.metal.device:
            api_token: "{{ apikey }}"
            project_id: "{{ project }}"
            name: "{{ name }}"
            state: present
            plan: "{{ plan }}"
            operating_system: "ubuntu_20_04"
            facility: "{{ region }}"
            wait_for_public_IPv: 6
            user_data: "{{ user_data }}"

2. Begin a de-provisioning of that server, either via the web UI or via setting state to absent on the above play.

3. Before the de-provisioning of the host created in Step 1 completes, re-run the above play.

EXPECTED RESULTS

Either the module to throw an error or to wait for the previous host to finish de-provisioning before attempting to re-provision.

ACTUAL RESULTS

The play succeeds, except output.devices is empty.