Segment deletion stuck in in-progress state due to orphaned DHCP logical port

Products

• VMware NSX

Issue/Introduction

• The deletion of a global network segment fails and throws a realization error because the management plane cannot remove the backing logical switch while a local internal DHCP/logical port attachment remains active.

• When deleting a segment from the UI, the process gets stuck in the In Progress status. The segment is greyed out and shows an In Progress status on both the Global and Local Managers. Clicking on the In Progress status reveals the following realization error popup window:

YYYY-MM-DDT4:30:00 PM: Unable to delete logical port with attachments of LogicalPort InternalLogicalPort/77969769-####-####-####-############.

• Logs Location: /var/log/proton/nsxapi.log

YYYY-MM-DDT23:10:00.297Z ERROR providerTaskExecutor-1-97 SwitchingValidator 4087112 SWITCHING [nsx@4413 comp="nsx-manager" errorCode="MP8414" level="ERROR" subcomp="manager"] The LogicalSwitch LogicalSwitch/a063d3df-####-####-####-############ already has dhcp_service attached on the LogicalPort InternalLogicalPort/-####-####-####-############, cannot create/update the LogicalPort SegmentPort/71232e1b-####-####-####-############.
YYYY-MM-DDT23:10:00.297Z  WARN providerTaskExecutor-1-97 DhcpAlarmHelper 4087112 POLICY [nsx@4413 comp="nsx-manager" level="WARNING" subcomp="manager"] com.vmware.nsx.management.switching.common.exceptions.SwitchingException: The LogicalSwitch LogicalSwitch/a063d3df-####-####-####-############ already has dhcp_service attached on the LogicalPort InternalLogicalPort/77969769-####-####-####-############, cannot create/update the LogicalPort SegmentPort/71232e1b-####-####-####-############.

Environment

• VMware Cloud Foundation
• VMware NSX

Cause

1. RealizedResourcePathUtil (mp/policy/lm-onboarding) independently re-derives DHCP realized-resource IDs using a plain <prefix>-<segmentId> formula.
2. The live DHCP provider names these resources <prefix>-<connectedGatewayId>-<segmentId> once the segment is attached to a gateway.
3. As the two ID schemes diverged, the onboarding's lookup for the "old" realized DHCP profile/server/logical-port/ip-pool/static-binding missed the exact stored object. 
4. Because of this condition, the  MigrationListenerImpl.deleteAndCreateRzResource() silently skipped the migration.

Resolution

Open a support request with Broadcom Support for assistance with safely and manually removing the orphaned DHCP internal logical port entry directly from the CorfuDB.