Customer Success Stories

Everything That Cost Us a Day Returned Success: Nine Versions of esx_ready.sh on a Live Fleet

Noah Farshad / October 23, 2026

One day, three fires, nine versions of esx_ready.sh on a live VCF 9.1.1 fleet, and nothing crashed: an upgrade that reported success and moved nothing, a cluster leave that released no disks, a depot staged onto the vSAN cache disk, a retried SDDC Manager task reading stale inventory. The build gate VCF 9.1.1 enforces, the chain of vSAN disk ownership, the feature that had to be deleted, and the stub harness that turned every failure into a test before its fix shipped.

The Certificates Expired a Year After a Date Nobody Was Watching: Five TKG Clusters Back From the Supervisor

Noah Farshad / October 23, 2026

Five VKS guest clusters on VCF 9.1.1, every API server down, every control plane certificate expired on the anniversary of a node rollout nobody remembered. Where TKG certificates actually live, why kubeadm’s renewal can hand you a certificate no client accepts, why the nodes have to be done in series when there is no quorum, and the script that took three production clusters from connection refused to READY in an afternoon.

Automation, Customer Success Stories

The NSX That Nobody Installed: Clearing Stale VIBs Before a VCF Import

Noah Farshad / October 16, 2026

A cluster would not convert to a vLCM image because of NSX nobody had installed: 48 VIBs in three layers, 23 of them invisible to vib list. How to find them by version, prove nothing is configured on top, and take them off with the command Broadcom supports.

Automation, Customer Success Stories

Prepared Is Not Working: The Overlay a VCF Import Leaves Out

Noah Farshad / October 16, 2026

Four imported clusters showed Prepared, Success and green hosts, and TEP IP Address: Not Set on every one. What a VCF import leaves behind, the five places an overlay breaks without an error, and a tool that copies the cluster that works.

Automation, Customer Success Stories

Importing ESXi 8 Into VCF 9.1 When ESX 9 Won’t Run on the Hardware

Noah Farshad / October 16, 2026

ESX 9 won’t run on Broadwell and VCF 9.1.1 won’t commission ESXi 8, so the clusters are built on vSphere 8 and imported. The vCenter upgrade that removed an NSX collision, the fourteen steps a cluster has to pass before VCF will take it, and the rules the tooling lives by.

Automation, Customer Success Stories

Avi 32.1.3 Under NSX Federation: One Controller Per Site, Local Objects Only

Noah Farshad / October 16, 2026

Avi 32.1.3 on VCF 9.1.1 with NSX Federation: controllers deployed through VCF Operations, a licence pool that was empty, a GM: prefix that stopped the first VIP, and KB 428430’s answer. One controller cluster per site, local objects only.

Automation, Customer Success Stories

Bridging Blue to Green: Moving Live Workloads to a New NSX Federation, One Subnet at a Time

Noah Farshad / October 16, 2026

Production VMs moved from a two-site NSX 4.2 Federation to a new VCF 9.1 Federation and kept their addresses: a Layer 2 bridge per subnet, waves of vMotions, a per-subnet gateway flip, and a tested way back from every step. The tooling is public.

Automation, Customer Success Stories

When the NSX Manager Certificate Is Already Expired

Noah Farshad / September 30, 2026

An expired NSX Manager certificate is replaced through the API: one certificate per node, the cluster VIP last, plan by default. The data plane is not touched.

Automation, Customer Success Stories

NSX Federation Made the Same Segment Appear Twice

Noah Farshad / September 30, 2026

Global Manager and Local Manager both show Aria Automation the same segment. Network profiles fill with duplicates, and IPAM does not always agree with either of them. Analyze, then plan, then execute.

Automation, Customer Success Stories

Sixteen Hosts, Seven Requirements, One Command: Getting ESXi Hosts Ready for VCF Commissioning

Noah Farshad / September 30, 2026

Two sites on VCF 9.1.1, one of them on hardware that cannot run ESX 9. VCF host commissioning has seven boring requirements and a wizard that reports one failure at a time; one POSIX shell script and one hosts.yaml took the fleet from racked to ready, upgrades, reboots and stale vSAN partitions included, refused the hosts that were not really standalone, and named the next command in every error.