Wingman — Remove

Destroy and verify disposable infrastructure

Resume the exact TF_VAR_run_id, private state directory, inventory, and observation journal created in Verify. Keep the shell environment from that run; never generate a new run ID for removal.

Check exact live ownership before planning and again before applying destruction. Stop if any UID changes; inspect the private state and journal rather than deleting a replacement. Do not run concurrent writers against these disposable names.

probe_root="$XCSH_WORKDIR/wingman/$TF_VAR_run_id"
python3 "$XCSH_WORKDIR/wingman/qualify.py" --verify-owned \
  --inventory "$probe_root/inventory.json" --private-directory "$probe_root/observations" \
  --kubeconfig "$XCSH_KUBECONFIG_FILE" --context "$XCSH_KUBECONFIG_CONTEXT" \
  --namespace "$XCSH_NAMESPACE"
terraform -chdir="$probe_root" plan -destroy -out=cleanup.plan
terraform -chdir="$probe_root" show cleanup.plan
shasum -a 256 "$probe_root/cleanup.plan"
python3 "$XCSH_WORKDIR/wingman/qualify.py" --verify-owned \
  --inventory "$probe_root/inventory.json" --private-directory "$probe_root/observations" \
  --kubeconfig "$XCSH_KUBECONFIG_FILE" --context "$XCSH_KUBECONFIG_CONTEXT" \
  --namespace "$XCSH_NAMESPACE"
terraform -chdir="$probe_root" apply cleanup.plan
python3 "$XCSH_WORKDIR/wingman/qualify.py" --verify-cleanup \
  --inventory "$probe_root/inventory.json" --private-directory "$probe_root/observations" \
  --kubeconfig "$XCSH_KUBECONFIG_FILE" --context "$XCSH_KUBECONFIG_CONTEXT" \
  --namespace "$XCSH_NAMESPACE"

Require successful Kubernetes reads showing no named Deployments and no run-labelled pods or ReplicaSets twice. Permission or transport errors do not count as absence. The empty create-state follow-up plan is checked before destroy; after destroy the state must contain no resources. Keep failed plan/apply evidence and reconcile state before making a fresh recovery plan.

Destroy your disposable qualification state through the saved cleanup plan, then verify Deployment, pod, and ReplicaSet absence twice. Revoke and remove your scoped kubeconfig after all needed checks. If you provisioned a reader environment, follow the operator teardown order to remove its route and runtime before bootstrap. Preserve the privately owned state and journals for recovery; never delete a retained deployment or sweep a namespace by prefix.

Detach the runtime origin

Set wingman.enabled = false in the frontend owner while retaining the integration inputs. Review and apply the listener update, then verify the Wingman route is absent and /get still works. Set wingman = null through a second reviewed plan to delete the detached origin pool. Destroy disposable runtime and bootstrap states only after frontend consumers are absent.