aaron/infra-core
1
0
Fork 1
Code Issues Pull requests 11 Projects Releases 1 Packages Wiki Activity Actions

HOMELAB-970: spec — SSD-tier migration design #181

Open
aaron wants to merge 1 commit from plane/HOMELAB-970-ssd-tier-migration-spec into live
pull from: plane/HOMELAB-970-ssd-tier-migration-spec
merge into: aaron:live
Conversation 0 Commits 1 Files changed 1 +214
aaron commented 2026-04-23 06:38:31 +00:00
Owner
Copy link

Summary

Design doc for moving 39 latency-sensitive PVCs onto a dedicated longhorn-fast StorageClass backed by ssd-mirror. HDD topology changes (re-mirror after sdb detach, add hdd-bulk pool, NFS media server, ARR stack) are explicitly deferred to follow-up tickets and are NOT part of this work stream.

Whats in the spec

  • Per-worker topology: 150 GB SSD-fast data disk per worker on ssd-mirror, tagged fast in Longhorn.
  • Keeps existing 1500 GB HDD virtio1 data disks attached on wk-01/02; wk-03 gets no HDD disk at all.
  • longhorn-fast SC (diskSelector=fast, 2 replicas, Retain) + edit longhorn-db to add diskSelector: "fast" so CNPG lands on SSD.
  • Keeps longhorn (default) untouched for backward compat.
  • 4-phase migration: infra → urgent 4 (already-full Prom TSDBs + ClickHouse + runner cache) → CNPG in-place switchovers → remaining 21 stateless → eventual HDD-tier decommission.
  • Fixes the audit finding that wk-02 Longhorn allowScheduling was false, causing all replicas to live on wk-01.
  • Capacity math: 75% ssd-mirror commitment post-migration (150×3 data + 264 GB boots).

Specialist agent reports (archived)

Spec built with 3 dispatched specialist agents:

  • PVC audit (full per-PVC inventory + tier recommendations)
  • ARR stack architect (the deferred media work, parked but referenced)
  • Longhorn storage architect (disk registration, SC, migration mechanics)

Reports at /tmp/dispatch-storage-audit-pvcs.md, /tmp/dispatch-arr-stack-plan.md, /tmp/dispatch-longhorn-tiered-design.md on the Mac.

Next steps (after this PR merges)

  1. Aaron reviews spec.
  2. Invoke superpowers:writing-plans skill to produce the step-by-step implementation plan.
  3. Implementation plan becomes a follow-up PR that edits the talos-cluster + proxmox-vm modules and the prod tfvars.

Test plan

  • Aaron reads the spec and confirms scope, phases, and capacity math.
  • Follow-up implementation plan ticket / PR tracks actual code changes.

🤖 Generated with Claude Code

## Summary Design doc for moving 39 latency-sensitive PVCs onto a dedicated `longhorn-fast` StorageClass backed by `ssd-mirror`. HDD topology changes (re-mirror after sdb detach, add hdd-bulk pool, NFS media server, ARR stack) are explicitly deferred to follow-up tickets and are NOT part of this work stream. ## Whats in the spec - Per-worker topology: 150 GB SSD-fast data disk per worker on ssd-mirror, tagged `fast` in Longhorn. - Keeps existing 1500 GB HDD virtio1 data disks attached on wk-01/02; wk-03 gets no HDD disk at all. - `longhorn-fast` SC (diskSelector=fast, 2 replicas, Retain) + edit `longhorn-db` to add `diskSelector: "fast"` so CNPG lands on SSD. - Keeps `longhorn` (default) untouched for backward compat. - 4-phase migration: infra → urgent 4 (already-full Prom TSDBs + ClickHouse + runner cache) → CNPG in-place switchovers → remaining 21 stateless → eventual HDD-tier decommission. - Fixes the audit finding that wk-02 Longhorn `allowScheduling` was false, causing all replicas to live on wk-01. - Capacity math: 75% ssd-mirror commitment post-migration (150×3 data + 264 GB boots). ## Specialist agent reports (archived) Spec built with 3 dispatched specialist agents: - PVC audit (full per-PVC inventory + tier recommendations) - ARR stack architect (the deferred media work, parked but referenced) - Longhorn storage architect (disk registration, SC, migration mechanics) Reports at `/tmp/dispatch-storage-audit-pvcs.md`, `/tmp/dispatch-arr-stack-plan.md`, `/tmp/dispatch-longhorn-tiered-design.md` on the Mac. ## Next steps (after this PR merges) 1. Aaron reviews spec. 2. Invoke `superpowers:writing-plans` skill to produce the step-by-step implementation plan. 3. Implementation plan becomes a follow-up PR that edits the talos-cluster + proxmox-vm modules and the prod tfvars. ## Test plan - [ ] Aaron reads the spec and confirms scope, phases, and capacity math. - [ ] Follow-up implementation plan ticket / PR tracks actual code changes. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
HOMELAB-970: spec — homelab SSD-tier migration design
Some checks failed
CI Review / pr-title (pull_request) Has been cancelled
Details
CI Review / helm-validate (pull_request) Has been cancelled
Details
CI Review / ai-review (pull_request) Has been cancelled
Details
Lint & Validate / terraform-validate (pull_request) Has been cancelled
Details
Lint & Validate / yaml-lint (pull_request) Has been cancelled
Details
Lint & Validate / shellcheck (pull_request) Has been cancelled
Details
1146272d8b 
Adds a dedicated longhorn-fast StorageClass backed by ssd-mirror (150 GB
SSD-fast data disk per worker) and migrates 39 latency-sensitive PVCs
off the current hdd-mirror-backed longhorn default.

Out of scope (deferred to follow-up tickets): HDD topology changes
(re-mirror, add hdd-bulk), NFS media server, ARR stack.

Designed with dispatched specialist agents (PVC audit, ARR architect,
Longhorn storage architect). Specialist reports archived at
/tmp/dispatch-*.md on the Mac for reference.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Some checks failed
CI Review / pr-title (pull_request) Has been cancelled
Details
CI Review / helm-validate (pull_request) Has been cancelled
Details
CI Review / ai-review (pull_request) Has been cancelled
Details
Lint & Validate / terraform-validate (pull_request) Has been cancelled
Details
Lint & Validate / yaml-lint (pull_request) Has been cancelled
Details
Lint & Validate / shellcheck (pull_request) Has been cancelled
Details
This pull request can be merged automatically.
This branch is out-of-date with the base branch
You are not authorized to merge this pull request.
View command line instructions

Checkout

From your project repository, check out a new branch and test the changes.
git fetch -u origin plane/HOMELAB-970-ssd-tier-migration-spec:plane/HOMELAB-970-ssd-tier-migration-spec
git switch plane/HOMELAB-970-ssd-tier-migration-spec
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
aaron/infra-core!181
No description provided.