## ✅Prerequisites
- VCF 9.0
- NSX Manager access
- vCenter access
## Problem
An NSX Edge node (`vcf-nsx-edge-01`) became non-responsive. Redeploying via NSX Manager (`System → Fabric → Nodes → Actions → Redeploy`) failed due to a CPU soft lockup on the edge node console during startup. The soft lockup was caused by performance issues documented in [[HowTo vSAN ESA Performance Intel X710 Workarounds VCF 9]].
## Resolution
### 1. Deploy a New Edge Node
Deployed a replacement edge node (`vcf-nsx-edge-03`) and migrated the T0 gateway interface to the new edge node.
### 2. Deletion of Old Edge Node Blocked
Deleting `vcf-nsx-edge-01` failed with Error 10026 — logical routers still referencing the old edge:
```
LogicalRouter/58c85c2b-567a-4744-a7f8-c33e9e2553c2
LogicalRouter/fbe8aa3f-67cc-4d4b-9f43-74d9ce5492b1
LogicalRouter/5d289ba0-67b1-4b9b-997c-70f0f9476b59
LogicalRouter/9ab5526a-898f-4e45-b9e0-2e5f9d9ace3a
LogicalRouter/659293ce-4ec2-4ac0-8257-e5f6a72e4ac0
LogicalRouter/35cbab68-e1d2-4996-a637-74c20cc3a39d
```
The vCenter Network Connectivity overview was also stuck at `100% Configuration in Progress`.
### 3. Unblock by Restarting vCenter
Restarting the vCenter appliance resolved the stuck state. After the restart, the old edge node could be deleted via vCenter under the Network Connectivity overview.
### Useful Troubleshooting Commands
Look up logical router references via NSX API:
```bash
GET https://<nsx-manager>/policy/api/v1/search?query=id:<uuid>
```
Check T0 locale services for stale edge references:
```bash
GET https://<nsx-manager>/policy/api/v1/infra/tier-0s/<t0-name>/locale-services
```
Check HA VIP, static routes, and BGP neighbors for old edge bindings.
## 🔗Resources
- [[HowTo NVMe Tiering AMD Ryzen Workaround VCF 9]]
- [[HowTo vSAN ESA Performance Intel X710 Workarounds VCF 9]]