Skip to main content

04 — Fleet analyze at scale

Goal: Score a directory of disks, find snowflakes and low-score blockers before wave planning.

1. Basic command​

export GUESTKIT_FLEET_JOBS=8
guestkit fleet analyze ./images/ | tee fleet.json
# or path layout your team uses for exports

Clusters OS fingerprints, flags outliers and low doctor scores. Use output to order waves (easy Linux first, BitLocker/domain Windows later) — or let fleet wave-plan derive the order from real dependency signals (DB role, NFS storage mounts) instead of doing it by hand:

guestkit fleet wave-plan ./images/ | tee waves.json

See migration-assurance.md.

TUI:

guestctl tui ./images/some.qcow2 --fleet ./images/
guestctl tui a.qcow2 --compare b.qcow2

2. Pipeline sketch​

  1. Export / sync wave N disks to scratch.
  2. fleet analyze → artifact fleet-wave-N.json.
  3. Ticket per blocker (score < floor, BitLocker, missing VirtIO plan).
  4. Per-disk repair loop (runbook 02) until Passport green.
  5. Only then schedule convert for that wave.

3. Capacity​

DisksGuidance
< 20Single worker, GUESTKIT_FLEET_JOBS=4
20–100Dedicated scratch SSD, jobs=8, stagger I/O
100+Shard directories by wave; multiple workers; object-store pull

Don’t run fleet analyze on the same spindles as live production storage.

4. Policy-as-code​

Combine fleet output with your org policy (required packages, forbidden domain-join at cutover, min score). Fail the wave gate if any disk remains below floor after repair SLA.

Related blog: Fleet inspect-batch.