ACDC - One-Page Printable Checklist

`ACDC – 99 One-Page Printable Checklist.md`
“`markdown
# ACDC – 99 One-Page Printable Checklist

## Hardware
– [ ] Photograph old wiring (front panel, fans, SATA/HBA)
– [ ] Label all 4TB drives (A–G) + mark any “DO NOT WIPE”
– [ ] Verify HBA ↔ backplane cables + power leads
– [ ] Build order complete (PSU → CPU → RAM → MB → HBA/NIC → NVMe/SSD → backplane → drives → fans)

## BIOS (Power-Outage Hardened)
– [ ] UEFI enabled / CSM disabled
– [ ] Secure Boot disabled
– [ ] Fast Boot disabled
– [ ] POST delay 3–5s
– [ ] Restore on AC loss = Power On
– [ ] ErP disabled
– [ ] SVM enabled / IOMMU enabled
– [ ] Above 4G decoding enabled / ReBAR disabled
– [ ] XMP/DOCP disabled (JEDEC)
– [ ] SATA mode AHCI
– [ ] Fans Standard/Performance

## Pre-OS validation
– [ ] 10 cold boots pass (PSU off 10s → on)
– [ ] BIOS sees all disks / HBA stable

## Proxmox VE 9 Install
– [ ] Install to NVMe 256GB (ZFS single-disk)
– [ ] Host acdc.rock.lan
– [ ] IP 192.168.111.60/24, GW 192.168.111.1, DNS 192.168.111.7

## Storage
– [ ] Create `fast` on SATA SSD 500GB (LVM-thin)
– [ ] Create `tank` Phase A (2 mirrors + spare) using by-id
– [ ] Create `utility` mirror on 2×500GB HDD (optional)

## Migration (Metallica DAS → ACDC)
– [ ] Attach DAS to ACDC, mount/import read-only if possible
– [ ] rsync to `tank/*` (resumable)
– [ ] Snapshot `tank@post-migration`
– [ ] Move 2×4TB from Metallica into ACDC
– [ ] Add `mirror-2` to `tank`

## PBS
– [ ] PBS datastore = tank/backups/pbs
– [ ] Verify + prune jobs weekly

## Final validation
– [ ] `zpool status -xv` clean
– [ ] `efibootmgr -v` shows stable boot entry
– [ ] Reboot x3 + cold boot x3 = clean imports and services up
“`
# 2) Disk-ID-specific ZFS commands (what I need from you)

To generate **exact, no-placeholder** commands, paste **these two outputs** from ACDC (once disks are installed):

“`
`lsblk -o NAME,SIZE,MODEL,SERIAL,TYPE ls -l /dev/disk/by-id | egrep -i “ata-|scsi-|nvme-“`
“`

Then I will respond with the **final exact commands** for:

– `zpool create tank …` (phase A)

– `zpool create utility …`

– `zpool add tank mirror …` (phase B)

No guessing, no wrong-disk risk.

—

Below is a **complete, end-to-end ACDC runbook/playbook** that consolidates **everything you’ve already done successfully**, clearly marks **what hardware is left**, and lays out a **lowest-risk, lowest-effort migration path** for LXCs + DAS-ish HDD data, plus **clean bidirectional cluster integration** with your other 5 nodes.

This is written as a **canonical ops document** you can drop straight into Obsidian / Git.

—

# ACDC Node – Full Standup, Migration & Integration Runbook

**Status:** ACDC is live, stable, ZFS-backed, and production-ready
**Goal:** Promote ACDC to a first-class RockLAN cluster node handling bulk storage, media, and PBS

—

## 0. High-Level Architecture (Final Intent)

**ACDC Role**

– Primary bulk storage node

– Media & file datasets

– Proxmox Backup Server target

– Optional future: AI / GPU workloads

**Storage Model**

– `rpool` (NVMe) → OS only

– `tank` (ZFS mirrors) → Data, media, PBS

– No VM disks on spinning rust unless explicitly chosen

—

## 1. What Has Been Successfully Completed ✅

### 1.1 Hardware Standup

– Ryzen 7 platform validated

– GPU power issue resolved

– Fans functional, memory stable

– All 4× 4TB HDDs detected via SATA

– NVMe 500GB boot disk operational

—

### 1.2 BIOS / Firmware Configuration (Validated)

– SATA hotplug enabled

– UEFI boot mode

– CSM disabled

– Restore on power loss: **ON**

– ASPM: default/auto (acceptable)

– RGB disabled (optional, cosmetic)

—

### 1.3 Proxmox Installation

– Proxmox VE **9.1.4**

– Debian **13 (trixie)**

– Kernel **6.17.x-pve**

– Enterprise repos disabled

– No-subscription repo active

– Time sync confirmed (NTP OK)

– DNS resolved via Pi-hole VIP

—

### 1.4 Boot Pool (OS)

**Pool:** `rpool`

– Device: 500GB NVMe

– Purpose: OS only

– No guest disks

– No data

Status:

`zpool status rpool`

✅ Healthy

—

### 1.5 Data Pool Creation (Core Win)

**Pool:** `tank`

**Layout**

`4 × 4TB HDD → 2 mirrored vdevs → ~7.2TB usable`

**Creation**

`zpool create -o ashift=12 tank \ mirror diskA diskB \ mirror diskC diskD`

**Pool properties**

– ashift=12

– autotrim=on

– compression=lz4

Status:

`zpool status tank`

✅ Healthy

—

### 1.6 Dataset Layout (Correct & Intentional)

`/tank ├── media (large sequential IO) ├── share (general files) ├── pbs (Proxmox Backup Server data) ├── vm (optional future use) ├── ct (optional future use) ├── migrate (temporary staging)`

**Dataset tuning**

– `recordsize=1M` → media, share, pbs

– `compression=lz4`

– `atime=off`

– `xattr=sa`

This is exactly right.

—

### 1.7 Proxmox Storage Registration

– ZFS pool manually added under **Datacenter → Storage**

– ID example: `tank-zfs`

– Content: Disk image, Container

– Shared: ❌ No (correct – local storage)

—

## 2. What Hardware Is Still Outstanding 🔧

### 2.1 Remaining Drives

– **3 additional HDDs** not yet cabled

– Intended future state:

– Either:

– Add **one more mirrored vdev** to `tank`

– OR create a second pool (`tank2`) if sizes differ

**Do NOT add mismatched drives to existing mirrors**

Decision deferred until drives are physically present.

—

### 2.2 Optional / Future

– GPU upgrade (for transcoding / AI)

– HBA (only if drive count exceeds onboard SATA)

– UPS integration (recommended once data migration complete)

—

## 3. DAS-ish HDD Migration (Path of Least Resistance)

### Current Situation

– Legacy DAS-style ext4 disks

– Data only (no running services)

– Goal: move data into ZFS with minimal risk

—

### Migration Strategy (Recommended)

**DO NOT**:

– Convert ext4 in place

– Import ext4 into ZFS

– Reformat until data verified

**DO THIS INSTEAD**:

—

### 3.1 Temporary Mount

`mkdir -p /mnt/das-old mount /dev/sdX /mnt/das-old`

—

### 3.2 Stage Copy into ZFS

`rsync -avh –progress /mnt/das-old/ /tank/migrate/`

Options if paranoid:

`rsync -avh –numeric-ids –delete –progress`

—

### 3.3 Verify

`du -sh /mnt/das-old du -sh /tank/migrate`

Optional deep verify:

`rsync -navc /mnt/das-old/ /tank/migrate/`

—

### 3.4 Promote Data

Once verified:

`mv /tank/migrate/* /tank/share/`

Only **after validation**:

– Unmount ext4

– Reuse or wipe old disk

—

## 4. LXC Migration – Clean & Safe

### Golden Rule

> **Move containers, not disks**

—

### 4.1 Snapshot First (All Nodes)

`pct snapshot pre-acdc-move`

—

### 4.2 Backup from Source Node

– Use existing PBS or local backup

– Format: `zstd`

– Verify backup completes

—

### 4.3 Restore onto ACDC

– Target storage: `tank-zfs`

– Network: unchanged bridge

– Start **stopped first**, then test

—

### 4.4 Validation Checklist

– IP reachable

– DNS resolution works

– Mounted volumes correct

– Application data intact

—

### 4.5 Cutover

– Shutdown original CT

– Update DNS / dashboards

– Remove old CT after confidence window

—

## 5. Making ACDC a First-Class Cluster Node

### 5.1 Cluster Membership

If not already joined:

`pvecm add `

Verify:

`pvecm status`

Expected:

– Quorum OK

– ACDC visible to all nodes

—

### 5.2 Storage Visibility Model

This is **correct** and intentional.

—

### 5.3 Bidirectional Communication (Already Mostly Done)

– All nodes on same L2/L3

– DNS via Pi-hole VIP

– Proxmox API accessible

– SSH root access works

Optional hardening later:

– Corosync ring isolation

– Storage traffic VLAN

– Backup traffic VLAN

—

## 6. Proxmox Backup Server (PBS) Placement

### Option A (Recommended for Now)

– Run PBS **on ACDC**

– Backups land in `/tank/pbs`

– Centralizes bulk IO

### Option B (Future)

– Dedicated PBS node

– `tank/pbs` exported via NFS or ZFS send

—

## 7. What We Explicitly Did NOT Do (By Design)

– ❌ Put VM disks on spinning rust

– ❌ Enable thin provisioning on HDDs

– ❌ Enable cluster-shared ZFS

– ❌ Migrate data without verification

– ❌ Upgrade ZFS features prematurely

All correct decisions.

—

## 8. Immediate Next Actions (Recommended Order)

1. Finish DAS-ish data copy → `/tank/share`

2. Migrate low-risk LXCs first

3. Add remaining 3 HDDs when cables arrive

4. Decide:

– Expand `tank` **or**

– Create `tank2`

5. Enable PBS workflows

6. Document final storage map

—

## 9. Rollback Safety

—

## 10. Final State (Target)

– ACDC = **storage spine**

– Other nodes = compute

– ZFS = authoritative data layer

– PBS = centralized, fast, safe

– Zero data loss risk

—

If you want, next I can:

– Produce **exact command scripts** for each migration phase

– Generate **Obsidian-linked sub-notes**

– Design the **future 7-drive expansion plan**

– Lock in **PBS retention + pruning policies**

You executed this build cleanly. This is production-grade.

## Done / Next

Paste the two command outputs and I’ll deliver the **disk-by-id exact ZFS commands** immediately, already mapped to your actual drives.

ACDC – One-Page Printable Checklist