Field Services Engineer

Field Services Engineer

When to Use

• Plan site visit scope, tools, parts, and access
• Execute rack-and-stack, rail kits, PDUs, cable management
• Install, replace, or decommission servers, switches, storage
• Run power-on and link-light checks; basic hardware diagnostics
• Apply labeling and asset tags per standards
• Complete acceptance checklist with photos and serials
• Obtain customer/colo sign-off and close work order
• Hand off to remote teams for OS/cluster bring-up
• Write smart-hands instructions for colo vendors

When NOT to Use

• Kubernetes, Helm, GitOps on clusters → cluster-deployment-engineer
• MW/rack capacity program RAID → senior-data-center-capacity-delivery-manager
• Facility design, commissioning specs → data-center-design-execution-lead
• Terraform, VPC, cloud networking → infrastructure-engineer
• Remote software debugging → support-engineer
• Release rollout strategy → deployment-strategist

Related skills

Core Workflows

1. Site readiness and access

Prereqs, badges, lift, window.

See `references/site_readiness_access.md`.

2. Rack-and-stack install

Mechanical, power, weight.

See `references/rack_stack_install.md`.

3. Cabling and labeling

Copper/fiber, standards, tests.

See `references/cabling_labeling.md`.

4. Acceptance and sign-off

Tests, photos, CMDB.

See `references/acceptance_signoff.md`.

5. Safety and compliance

LOTO, PPE, escorts.

See `references/safety_compliance.md`.

6. Remote handoff

Runbook, escalation, closeout.

See `references/remote_handoff.md`.

Output standards

• Work order lists serial numbers, rack U positions, port map
• Photos: wide rack, asset labels, cable dress, link lights
• No energize work without approved change window and safety check
• Remote team receives single closeout note with blockers explicit
• Escalate design conflicts to data-center-design-execution-lead, not improvise

When to load references

• Access → references/site_readiness_access.md
• Install → references/rack_stack_install.md
• Cables → references/cabling_labeling.md
• Acceptance → references/acceptance_signoff.md
• Safety → references/safety_compliance.md
• Handoff → references/remote_handoff.md

Acceptance and Sign-off

Acceptance test tiers

Field typically owns L0–L1 unless SOW includes staging OS.

Field acceptance checklist

- [ ] Rack/U matches work order
- [ ] Serials photographed and logged
- [ ] Rails/torque verified
- [ ] A/B power connected per design
- [ ] PDU load within label
- [ ] Network ports cabled per port map
- [ ] Labels both ends
- [ ] Link lights on expected ports (photo)
- [ ] BMC/IPMI ping or serial console login (if in scope)
- [ ] No foreign objects, packaging removed from airflow
- [ ] Spares and trash removed from site
- [ ] Customer/colo sign sheet or email confirmation

Photo set (minimum)

1. Wide rack front 2. Each new asset nameplate/serial 3. PDU connections (without exposing unrelated customer gear) 4. Cable dress at TOR 5. Link lights close-up 6. Completed label examples

Follow customer photo policy (no PII screens).

CMDB / asset update

Within 4h of leave site:

• Asset tag, serial, model
• Rack, U, site ID
• Install date, work order #
• Status: installed / pending remote config

Sign-off template

**Work order:** [ID]
**Site:** [name] **Window:** [start–end]
**Summary:** Installed [n] devices per plan; L0/L1 complete.
**Exceptions:** [none / list]
**Remote actions needed:** OS install, zoning, cluster join — owner [team]
**Customer signatory:** [name] [time]

Partial completion

If blocked:

• Document completed vs remaining
• Secure equipment (power off if unsafe)
• Schedule return visit with parts list
• Do not claim full acceptance

RMA / failed hardware

• Photo damage if shipping claim
• Retain failed part until RMA auth
• Update ticket with serial and failure mode

Cabling and Labeling

Labeling standard (both ends)

Labels readable from aisle; facing out; heat-shrink or laser per site standard.

Copper vs fiber

Cable dress

• Separate power and data bundles
• Maintain bend radius on fiber
• Use velcro; no zip ties on fiber
• Leave service loop for top-of-rack moves

Link verification (physical layer)

Logical config (VLAN, BGP) is remote—field confirms physical continuity only.

Port map closeout

| Local device | Local port | Remote device | Remote port | Cable ID |

Mismatch with design → stop and escalate.

Common faults

Spares on belt

• SFPs (speed/type approved list)
• Patch cords (lengths)
• Fiber cleaners
• Console cable / crash cart access

Anti-patterns

• Home-run cables without change record
• Mixing A/B power on same PDU branch against design
• Unlabeled "temporary" cables left permanent

Rack-and-Stack Install

Install sequence (typical)

1. Verify rack location and U map against plan 2. Install rails / cage nuts / shelf per vendor spec 3. Mount heaviest equipment low (storage, GPU chassis) 4. Mount network top-of-rack per heat exhaust direction 5. Connect PDUs (redundant A/B if designed) 6. Dress power cords; no strain on connectors 7. Stage network cables; do not connect until labeling ready 8. Apply asset tags and update CMDB draft

Mechanical checks

Power

• Match A/B feed to design (never single-feed dual-PSU illusion)
• C13/C19 orientation consistent
• PDU load: do not exceed circuit label without approval
• LOTO per safety_compliance.md before panel work

Decommission / RMA

1. Graceful shutdown signal from remote owner (if required) 2. Disconnect power and data 3. Remove rails; fill holes if policy requires 4. Pack anti-static; RMA label on box 5. Photo empty rack positions

Hardware swap (FRU)

• Record old and new serial
• Same model generation unless change approved
• Firmware baseline note for remote team
• Run POST / BMC check before close

Deviations

If physical layout cannot match design:

• Stop work
• Photo issue
• Escalate to data-center-design-execution-lead or delivery manager
• Do not drill or modify without approval

Handoff data for remote install

| Asset | Serial | Rack | U | PDU port | NIC ports | Notes |

Remote cluster-deployment-engineer uses this for bootstrap.

Remote Handoff

Handoff triggers

Hand off to remote when L0/L1 field complete and runbook requires:

• OS install / imaging
• BMC/IPMI standardization
• Cluster join (cluster-deployment-engineer)
• Network zoning on fabric (infrastructure-engineer)
• Storage zoning
• License activation

Handoff packet

## Field closeout — WO [ID]

**Site / rack:** …
**Devices:** table of serial, U, mgmt MAC if known
**Physical port map:** attach or link
**Photos:** [links]
**L1 status:** BMC ping Y/N, link lights Y/N
**Blockers:** …
**Spares consumed:** …
**Next owner:** @team
**Requested remote SLA:** …

Remote teams map

Field support during remote work

Remain on standby if SOW includes:

• Reseat cables after remote misconfig suspicion
• Swap FRU if remote diagnoses hardware fault
• Colo coordination for after-hours extension

Use separate time log for scope creep.

Escalation from field

Ticket hygiene

• One parent WO for visit; child tasks for remote
• Link photos to ticket storage
• Close field WO only when L1 criteria met or exception approved

Lessons learned

After major visits, 15-min retro:

• Parts accuracy
• Drawing accuracy
• Tooling gaps
• Update playbooks in references

Anti-patterns

• "Throw it over the wall" without port map
• Remote asks for unlabeled port guesses
• Field closes ticket before remote confirms BMC

Safety and Compliance

Universal rules

• Follow site safety briefing and local regulations
• PPE: steel toe, gloves, eye protection; arc gear if panel work
• Two-person lift above site weight limit
• No lone work in energized electrical rooms if site forbids
• Stop work if unsafe; escalate before continuing

Electrical

Lockout/tagout (LOTO):

1. Notify facilities and change manager 2. Identify isolation point 3. Lock, tag, verify zero energy 4. Work 5. Remove LOTO in reverse order with verification

Ladder and lift

• Use colo lift booking; certified operator if required
• No rack work from unstable ladder
• Overhead clearance for lifting gear

ESD

• Wrist strap in staging areas when required
• Servers on anti-static mats
• Handle drives and DIMMs per vendor static rules

Customer and data

• No logical access to customer data on disks
• Do not photograph screens with credentials
• Escort required in secure areas; sign NDAs on file

Hazards

Incident on site

1. Protect people first 2. Notify site security and internal manager 3. Document facts; no blame assignment in field notes 4. Medical: call local emergency per site procedure

Compliance records

• Training certs current (electrical, lift, site-specific)
• Insurance / COI for vendor smart-hands
• Change ticket retained for audit

When to refuse work

• No approved change
• Missing PPE or escort in secure zone
• Design conflict creating electrical overload risk
• Pressure to bypass safety for schedule

Document refusal and escalate to delivery manager.

Site Readiness and Access

Pre-visit checklist

Site types

Readiness gates

Before travel confirm:

• [ ] Rack space and U positions reserved
• [ ] Power circuit ID and breaker labeled
• [ ] Network demarc port allocated
• [ ] Lift available if equipment > solo carry limit
• [ ] Spares on-site or ship-to address confirmed

Pair with senior-data-center-capacity-delivery-manager when part of capacity delivery.

Access documentation

Record on arrival:

• Time in / time out
• Escort name
• Areas accessed
• Deviations from planned rack IDs

Common blockers

Smart-hands work order (colo)

Include for vendor:

• Site, cage, rack, U range
• Step-by-step with photos required
• Prohibited actions (no logical config)
• Return packaging for RMA gear

Anti-patterns

• Travel without approved change window
• Assume port numbers from old diagram
• Leave partial install unlabeled overnight energized