“Anyone can deploy a lab. The true mark of engineering discipline is building an automated teardown workflow that reclaims 100% of storage, memory, and IP addresses.”

When managing multi-tenant SDDC test pods at NTT Data, engineers frequently finished testing and abandoned running virtual machines.

Over time, orphaned VMDK disk files consumed 1TB of NVMe datastore capacity, while stale BGP neighbor entries remained active on corporate transit switches.

Engineering the undeploy.yml Teardown Pipeline

We engineered an automated undeploy.yml playbook that executes reverse tear-down sequence:

1. Graceful Appliance Shutdown: Initiates soft shutdown of nested ESXi hosts and vCenter to prevent file system corruption.
2. BGP Session Teardown: Administrative withdrawal of BGP routes from physical switches.
3. Storage & DNS Purge: Unregisters VM objects, deletes datastore folders, and removes PTR/A DNS records via API.

[!IMPORTANT] Always enforce automated TTL (Time-To-Live) expiration timers on test environments to trigger mandatory teardown playbooks after 72 hours.

# # Ansible Automated Teardown Task Snippet (undeploy.yml)
- name: "Gracefully Power Off and Delete Pod VMs"
  community.vmware.vmware_guest:
    hostname: "{{ physical_esxi }}"
    username: "{{ user }}"
    password: "{{ pass }}"
    name: "{{ item }}"
    state: absent
    force: yes
  loop:
    - "{{ pod_prefix }}-NSXT-Edge02"
    - "{{ pod_prefix }}-NSXT-Edge01"
    - "{{ pod_prefix }}-NSXT-LM"
    - "{{ pod_prefix }}-vCenter"
    - "{{ pod_prefix }}-ESXi03"
    - "{{ pod_prefix }}-ESXi02"
    - "{{ pod_prefix }}-ESXi01"

The Verdict

Key Takeaway

Build Automated Teardown Playbooks alongside Provisioning Code.

Treat lab destruction as a first-class engineering requirement to prevent datastore exhaustion, IP address leakage, and zombie cloud infrastructure.