Overview — Proforma Infrastructure Engineer Interview
The interview, at a glance
- Role: Infrastructure Engineer — primary technical owner of the container networking stack
- Company: Proforma — $650M print/promo/packaging distributor, Tampa (Davis Islands)
- Date: Wednesday, September 16, 2026 — 1 hour, single interviewer (hiring manager)
- Comp: $100k–$120k
- Team context: two-person team today — one person owns the work. They are hiring to remove a single point of failure. Every answer you give should quietly show you can BE the redundancy.
- Their stack: hybrid Windows/Linux + OpenShift, MetalLB, OVN-Kubernetes, BGP, micro-segmentation, RHEL, Windows Server + AD, on-prem.
The one sentence that wins the interview
> "You're hiring someone to own the container networking stack — BGP, MetalLB, OVN-K — plus the Windows-to-OpenShift migration. That's my current job: I run BGP on the F5s and the OpenShift edge, I run OpenShift for Java microservices, and I've moved Linux workloads between platforms for years. I can walk you through each layer."
The job posting itself conflates "OpenShift Virtualization (OVN-Kubernetes)". Splitting those two cleanly — OpenShift Virtualization = KubeVirt (running VMs), OVN-Kubernetes = the CNI (cluster networking) — is an instant credibility win. Say it once, early.
Your narrative spine
One story told four ways, so you never contradict yourself:
- Topology — Internet → core switches (BGP lives here) → F5 BIG-IP pair (VIPs) → OpenShift cluster (masters + workers) → OpenShift Virtualization (migrated Windows VMs). Four boxes. Draw it on the Whiteboard page.
- BGP — F5s peer eBGP into the core and advertise the VIP subnets, only from the active unit. Same reachability problem MetalLB solves at the cluster edge: speakers announce /32s, fabric ECMPs to the workers.
- OpenShift — masters, workers, OVN-K overlay, microservices, namespaces, RBAC, NetworkPolicies.
- Migration — Windows VMs leaving VMware enter the cluster via OpenShift Virtualization; VMware knowledge (V2V, VxRail, PowerFlex) is your credibility for the migration project.
Study sequence — follow this order
| Order | Page | What it's for | Time |
|---|---|---|---|
| 1 | Company & Role | Who you're talking to, what they need | 10 min |
| 2 | BGP Deep Dive | The technical core of the role | 45 min |
| 3 | OpenShift + OVN + MetalLB | Their home turf — know it cold | 45 min |
| 4 | Windows VM Migration | The project they're staffing | 30 min |
| 5 | L4 vs L7 | F5 credibility, fast read | 15 min |
| 6 | Hardening | The second pillar of the role | 25 min |
| 7 | Mock Interview Script | All 35 questions, in order, with what he's probing | 60 min |
| 8 | Q&A Bank | Model answers + your STAR stories + questions to ask him | 60 min |
| 9 | Whiteboard | The diagram + narration + live-draw rules | 30 min |
Total first pass: ~5 hours. You have today (Sunday) and Monday evening. Do one full pass, then re-drill BGP + OpenShift + Whiteboard Tuesday night.
The 2-day plan
Today / tomorrow (content):
- BGP page (02) — read it twice. This is where the interview is won or lost.
- OpenShift page (03) — twice. You must speak OVN-K and MetalLB fluently.
- Company page (01) + Q&A bank (07) — your stories and the "why us" angle.
- Whiteboard page (09) — draw it from memory until it takes 2 minutes.
Tuesday night (drill):
- Mock script (08) — answer every question OUT LOUD, one at a time. Not in your head.
- The whiteboard, five times. Narrate each box out loud.
Wednesday morning (light):
- Skim L4 vs L7 (05) + Hardening (06) — refresh, don't cram.
- Re-read the one sentence above and your four stories.
Ground-truth at work — 10 minutes, before Wednesday (non-negotiable):
The difference between "probably a mix" and a real answer is the specifics. While you still have Penske access:
F5: tmsh list net routing bgp
imish -e 'show ip bgp summary'
OpenShift: oc get ipaddresspool,bgppeer,bgpadv -A
oc -n metallb-system get pods
Capture: your real ASN, neighbor IPs, VIP subnets, worker count, OCP version. Write them onto your whiteboard diagram. Those specifics are what separate you from the next candidate.
Traps to avoid
- Don't contradict yourself. Every answer must fit the four-box narrative. If a detail doesn't fit, leave it out.
- Don't fake depth. If he goes past your depth, say "I haven't run that in production, here's how I'd approach it" — that answer is respected; invented specifics are not.
- NinjaOne (their RMM). You've used CrowdStrike for 1,000+ endpoints and managed patching with Ansible. The honest mapping: "I haven't run NinjaOne, but it's the same problem space as what I've done — deployment, patching, monitoring. I'd be up to speed in a week." Do NOT claim hands-on NinjaOne.
- MetalLB question. They will ask about it (it's in the posting). Your line: "Same pattern, different edge — MetalLB speakers on the workers announce /32s into the fabric." One sentence, then keep moving.
- No invented features or claims — on the site, in the interview, anywhere. Placeholder specifics (ASN, VIP ranges) stay blank until you ground-truth them.
Day-of checklist
- 10 min before: paper + pen next to you (whiteboard fallback), water, this site open.
- First 2 min: the one sentence, delivered calmly.
- Interruptions: he will interrupt. That's engagement. Answer in one sentence, return to your thread.
- End: three questions for him (see Q&A Bank — pick your three and memorize them).
- After: note every question he asked that surprised you. They'll matter if there's a round two.