fix(k3s): route audit logs through journald, codify etcd member recovery

Two changes prompted by today's etcd raft panic on worker1-k8s0
(tocommit out of range, lost-write on follower) and the cascading
disk pressure that surfaced underneath it.

Audit logs to journald
- kube-apiserver now uses audit-log-path=- so audit events flow to
  k3s.service stdout and into journald instead of growing files in
  /var/log/kubernetes. The previous setup combined apiserver's
  internal rotation with a logrotate *.log glob that double-rotated
  the rotated files into permanent orphans (observed: 7+ GB).
- New journald-limits operation writes a SystemMaxUse=2G drop-in so
  audit volume cannot fill /var/log even under bursty load.
- log-rotation operation repurposed to decommission the obsolete
  logrotate rule and reap leftover audit files. Idempotent: no-op
  on fresh installs.

Etcd member recovery
- New recoverEtcdMember(broken, peer, hostname) codifies the
  documented k3s recovery: stop k3s, etcdctl member remove, wipe
  /var/lib/rancher/k3s/server/{db,tls,cred}, restart, poll for
  rejoin. Refuses to operate when cluster size < 3 to preserve
  quorum.

Tests
- 7 new unit tests covering both decommission paths and the
  recovery procedure (54 total, all green).
- install.test.ts asserts the file-based audit args are gone.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.gitignore

@@ -27,3 +27,7 @@ node_modules/
 # Task files
 # tasks.json
 # tasks/ 
+
+# Asahi build artifacts (large)
+bastion/.asahi-cache/
+bastion/asahi-repo/*.zip

CLAUDE.md

@@ -0,0 +1,19 @@
+## Skill routing
+
+When the user's request matches an available skill, ALWAYS invoke it using the Skill
+tool as your FIRST action. Do NOT answer directly, do NOT use other tools first.
+The skill has specialized workflows that produce better results than ad-hoc answers.
+
+Key routing rules:
+- Product ideas, "is this worth building", brainstorming → invoke gstack-office-hours
+- Bugs, errors, "why is this broken", 500 errors → invoke gstack-investigate
+- Ship, deploy, push, create PR → invoke gstack-ship
+- QA, test the site, find bugs → invoke gstack-qa
+- Code review, check my diff → invoke gstack-review
+- Update docs after shipping → invoke gstack-document-release
+- Weekly retro → invoke gstack-retro
+- Design system, brand → invoke gstack-design-consultation
+- Visual audit, design polish → invoke gstack-design-review
+- Architecture review → invoke gstack-plan-eng-review
+- Save progress, checkpoint, resume → invoke gstack-checkpoint
+- Code quality, health check → invoke gstack-health

TODOS.md

@@ -0,0 +1,47 @@
+# TODOS
+
+## P1 — Ship with Phase 1
+
+### v2.0 Architecture Document Update
+Update `bastion/docs/ARCHITECTURE.md` to cover v2.0: driver model, fleet system,
+Pulumi integration, Vault secrets, Deno evaluator, new CLI grammar. The existing
+doc covers v1.0 comprehensively (432 lines). v2.0 adds 5+ major subsystems.
+**Effort:** M (human: 1 week / CC: 1-2 days)
+**Depends on:** Phase 1 complete
+**Source:** CEO review 2026-04-01
+
+## P2 — Post-v2.0 Core
+
+### SSH Emergency Mode (scoped)
+SSH-based operations limited to: (1) earliest necessary box provisioning before agent
+is installed, and (2) emergency debugging/fixing operations that can't be done via agent.
+NOT a general-purpose DeploymentTarget alternative. The v1.0 `recheck` and `fix-ssh-root.sh`
+patterns are the model. Agent stays the primary management path.
+**Effort:** S (human: 1 week / CC: 1 day)
+**Depends on:** Phase 2 complete (DeploymentTarget interface exists)
+**Source:** CEO review 2026-04-01
+
+### Prometheus Metrics Endpoint
+Add `/metrics` endpoint to labd: resource counts by status, apply duration histograms,
+driver operation latency, fleet pipeline completion rates. Standard Prometheus scraping
+for Grafana dashboards and alerting.
+**Effort:** S (human: 2-3 days / CC: 2-3 hours)
+**Depends on:** Phase 1 (labd exists with resource store)
+**Source:** CEO review 2026-04-01 (observability gap)
+
+## P3 — Future Enhancements
+
+### Infrastructure Graph Visualization
+Visual representation of resource dependencies, environment topology, fleet status.
+Could be a web UI or terminal-based (like `kubectl tree`).
+**Source:** CEO review 2026-04-01
+
+### `labctl import` for Existing Cloud Resources
+Discover and import existing AWS/GCP resources into the state store.
+Pulumi's import functionality could be leveraged.
+**Source:** CEO review 2026-04-01
+
+### Built-in Secrets Rotation
+Automatic rotation of managed secrets (database passwords, API keys).
+Vault handles rotation but a labctl-native workflow could simplify.
+**Source:** CEO review 2026-04-01

bastion/asahi-repo/installer_data.json

@@ -0,0 +1,47 @@
+{
+  "os_list": [
+    {
+      "name": "Fedora Asahi Lab (infra)",
+      "default_os_name": "Fedora Linux Lab",
+      "boot_object": "m1n1.bin",
+      "next_object": "m1n1/boot.bin",
+      "package": "fedora-asahi-lab.zip",
+      "supported_fw": [
+        "12.3",
+        "12.3.1",
+        "13.5"
+      ],
+      "partitions": [
+        {
+          "name": "EFI",
+          "type": "EFI",
+          "size": "524288000B",
+          "format": "fat",
+          "volume_id": "0x804be8a6",
+          "copy_firmware": true,
+          "copy_installer_data": true,
+          "source": "esp"
+        },
+        {
+          "name": "Boot",
+          "type": "Linux",
+          "size": "1073741824B",
+          "image": "boot.img"
+        },
+        {
+          "name": "Root",
+          "type": "Linux",
+          "size": "4626296832B",
+          "expand": false,
+          "image": "root.img"
+        },
+        {
+          "name": "Data",
+          "type": "Linux",
+          "size": "1073741824B",
+          "expand": true
+        }
+      ]
+    }
+  ]
+}

bastion/bastion/.gitignore

@@ -0,0 +1,4 @@
+
+# Asahi build artifacts (large)
+.asahi-cache/
+asahi-repo/*.zip

bastion/completions/labctl.bash

@@ -73,12 +73,18 @@ _labctl() {
     "provision register")
       COMPREPLY=($(compgen -W "--role --ip -h --help" -- "$cur"))
       return ;;
+    "provision asahi")
+      COMPREPLY=($(compgen -W "-h --help" -- "$cur"))
+      return ;;
     "provision logs")
       COMPREPLY=($(compgen -W "-f --follow -h --help" -- "$cur"))
       return ;;
     "provision makeiso")
       COMPREPLY=($(compgen -W "--arch --local --out -h --help" -- "$cur"))
       return ;;
+    "provision recheck")
+      COMPREPLY=($(compgen -W "--user --target -h --help" -- "$cur"))
+      return ;;
     "config list")
       COMPREPLY=($(compgen -W "-h --help" -- "$cur"))
       return ;;
@@ -104,7 +110,7 @@ _labctl() {
       COMPREPLY=($(compgen -W "bastion -h --help" -- "$cur"))
       return ;;
     "provision")
-      COMPREPLY=($(compgen -W "list install reprovision debug forget register logs makeiso -h --help" -- "$cur"))
+      COMPREPLY=($(compgen -W "list install reprovision debug forget register asahi logs makeiso recheck -h --help" -- "$cur"))
       return ;;
     "config")
       COMPREPLY=($(compgen -W "list get set path -h --help" -- "$cur"))

bastion/completions/labctl.fish

@@ -125,8 +125,10 @@ complete -c labctl -n "__labctl_using_cmd provision" -a reprovision -d 'Queue in
 complete -c labctl -n "__labctl_using_cmd provision" -a debug -d 'PXE boot into Fedora rescue mode for debugging (target: hostname, MAC, or IP)'
 complete -c labctl -n "__labctl_using_cmd provision" -a forget -d 'Remove a machine from bastion state'
 complete -c labctl -n "__labctl_using_cmd provision" -a register -d 'Register an already-installed machine (e.g. after state loss)'
+complete -c labctl -n "__labctl_using_cmd provision" -a asahi -d 'Show instructions to provision an Apple Silicon Mac with Asahi Linux'
 complete -c labctl -n "__labctl_using_cmd provision" -a logs -d 'Show provisioning logs for a machine (hostname, MAC, or IP)'
 complete -c labctl -n "__labctl_using_cmd provision" -a makeiso -d 'Generate a UEFI-bootable iPXE ISO for network provisioning'
+complete -c labctl -n "__labctl_using_cmd provision" -a recheck -d 'Refresh hardware info for all installed machines via SSH'
 
 # provision install options
 complete -c labctl -n "__labctl_in_cmd provision install" -l role -d 'Machine role (see below)' -xa 'vanilla worker infra labcontroller'
@@ -153,6 +155,10 @@ complete -c labctl -n "__labctl_in_cmd provision makeiso" -l arch -d 'Target arc
 complete -c labctl -n "__labctl_in_cmd provision makeiso" -l local -d 'Build ISO locally instead of using bastion-hosted URL'
 complete -c labctl -n "__labctl_in_cmd provision makeiso" -l out -d 'Output path for local ISO build' -x
 
+# provision recheck options
+complete -c labctl -n "__labctl_in_cmd provision recheck" -l user -d 'SSH user' -x
+complete -c labctl -n "__labctl_in_cmd provision recheck" -l target -d 'Only recheck a specific machine (by hostname or MAC)' -x
+
 # config subcommands
 complete -c labctl -n "__labctl_using_cmd config" -a list -d 'Show all configuration values'
 complete -c labctl -n "__labctl_using_cmd config" -a get -d 'Get a configuration value'

bastion/package.json

@@ -22,7 +22,11 @@
     "test:integration:iso": "vitest run -c tests/integration/vitest.config.ts -t 'ISO boot'",
     "test:integration:iso:host": "sudo -E $(which npx) vitest run -c tests/integration/vitest.config.ts -t 'ISO boot'",
     "test:integration:arm-iso": "vitest run -c tests/integration/vitest.config.ts -t 'ARM ISO'",
-    "test:integration:arm-iso:host": "sudo -E $(which npx) vitest run -c tests/integration/vitest.config.ts -t 'ARM ISO'"
+    "test:integration:arm-iso:host": "sudo -E $(which npx) vitest run -c tests/integration/vitest.config.ts -t 'ARM ISO'",
+    "test:integration:asahi": "vitest run -c tests/integration/vitest.config.ts -t 'asahi firstboot'",
+    "test:integration:asahi:host": "sudo -E $(which npx) vitest run -c tests/integration/vitest.config.ts -t 'asahi firstboot'",
+    "test:integration:asahi-validate": "vitest run -c tests/integration/vitest.config.ts -t 'asahi.*validation'",
+    "test:integration:asahi-validate:host": "sudo -E $(which npx) vitest run -c tests/integration/vitest.config.ts -t 'asahi.*validation'"
   },
   "engines": {
     "node": ">=20.0.0",

bastion/scripts/build-asahi-rootfs.sh

@@ -0,0 +1,302 @@
+#!/bin/bash
+# Build a custom Fedora Asahi Remix rootfs with lab firstboot LVM setup.
+#
+# Downloads the upstream Fedora Asahi Remix Server package, injects our
+# firstboot script + systemd service, and repackages it for the bastion.
+#
+# Requirements: root, curl, unzip, mount (loop), zip
+# Output: bastion/asahi-repo/ directory with package + installer_data.json
+#
+# Usage: sudo ./scripts/build-asahi-rootfs.sh [--bastion-ip IP] [--http-port PORT]
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+PROJECT_DIR="$(cd "$SCRIPT_DIR/.." && pwd)"
+ASAHI_DIR="$PROJECT_DIR/asahi-repo"
+CACHE_DIR="$PROJECT_DIR/.asahi-cache"
+WORK_DIR=""
+
+# Defaults
+BASTION_IP="${BASTION_IP:-192.168.8.23}"
+HTTP_PORT="${HTTP_PORT:-8080}"
+ROLE="${ROLE:-infra}"
+HOSTNAME="${HOSTNAME:-mac-studio}"
+MAC="${MAC:-00:00:00:00:00:00}"
+ADMIN_USER="${ADMIN_USER:-michal}"
+
+# Parse args
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --bastion-ip) BASTION_IP="$2"; shift 2 ;;
+        --http-port) HTTP_PORT="$2"; shift 2 ;;
+        --role) ROLE="$2"; shift 2 ;;
+        --hostname) HOSTNAME="$2"; shift 2 ;;
+        --mac) MAC="$2"; shift 2 ;;
+        --admin-user) ADMIN_USER="$2"; shift 2 ;;
+        *) echo "Unknown option: $1"; exit 1 ;;
+    esac
+done
+
+# ── Resolve upstream package URL ─────────────────────────────────
+echo "==> Fetching Asahi installer data..."
+INSTALLER_DATA=$(curl -sfL "https://cdn.asahilinux.org/installer/installer_data.json")
+
+# Find the Server variant package URL
+SERVER_URL=$(echo "$INSTALLER_DATA" | python3 -c "
+import sys, json
+data = json.load(sys.stdin)
+for os in data.get('os_list', []):
+    name = os.get('name', '').lower()
+    if 'server' in name and 'uefi' not in name and not os.get('expert'):
+        print(os['package'])
+        break
+" 2>/dev/null)
+
+if [ -z "$SERVER_URL" ]; then
+    echo "ERROR: Could not find Fedora Asahi Remix Server in installer data."
+    echo "Available variants:"
+    echo "$INSTALLER_DATA" | python3 -c "
+import sys, json
+data = json.load(sys.stdin)
+for os in data.get('os_list', []):
+    print(f\"  - {os.get('name', '?')}\")" 2>/dev/null
+    exit 1
+fi
+
+PACKAGE_NAME=$(basename "$SERVER_URL")
+echo "    Variant: Fedora Asahi Remix Server"
+echo "    Package: $PACKAGE_NAME"
+
+# Also extract the partition layout and supported_fw from upstream
+UPSTREAM_CONFIG=$(echo "$INSTALLER_DATA" | python3 -c "
+import sys, json
+data = json.load(sys.stdin)
+for os in data.get('os_list', []):
+    name = os.get('name', '').lower()
+    if 'server' in name and 'uefi' not in name and not os.get('expert'):
+        json.dump(os, sys.stdout)
+        break
+")
+
+# ── Download upstream package ────────────────────────────────────
+mkdir -p "$CACHE_DIR" "$ASAHI_DIR"
+
+CACHED_PKG="$CACHE_DIR/$PACKAGE_NAME"
+if [ -f "$CACHED_PKG" ]; then
+    echo "==> Using cached package: $CACHED_PKG"
+else
+    echo "==> Downloading $SERVER_URL..."
+    curl -# -L -o "$CACHED_PKG" "$SERVER_URL"
+fi
+
+# ── Extract and modify rootfs ────────────────────────────────────
+WORK_DIR=$(mktemp -d)
+trap 'echo "==> Cleaning up..."; umount "$WORK_DIR/rootfs" 2>/dev/null || true; rm -rf "$WORK_DIR"' EXIT
+
+echo "==> Extracting package..."
+unzip -q -o "$CACHED_PKG" -d "$WORK_DIR/pkg"
+
+# List contents
+echo "    Package contents:"
+ls -lh "$WORK_DIR/pkg/" | grep -v ^total | while read -r line; do echo "      $line"; done
+
+# Find root.img
+ROOT_IMG=$(find "$WORK_DIR/pkg" -name "root.img" -type f | head -1)
+if [ -z "$ROOT_IMG" ]; then
+    echo "ERROR: root.img not found in package."
+    echo "Contents: $(ls "$WORK_DIR/pkg/")"
+    exit 1
+fi
+
+echo "==> Mounting root.img..."
+mkdir -p "$WORK_DIR/rootfs"
+mount -o loop "$ROOT_IMG" "$WORK_DIR/rootfs"
+
+# ── Read SSH keys from the system ────────────────────────────────
+SSH_KEYS=""
+REAL_USER="${SUDO_USER:-$USER}"
+REAL_HOME=$(eval echo "~$REAL_USER")
+for keyfile in "$REAL_HOME/.ssh/id_ed25519.pub" "$REAL_HOME/.ssh/id_ecdsa.pub" "$REAL_HOME/.ssh/id_rsa.pub"; do
+    if [ -f "$keyfile" ]; then
+        SSH_KEYS=$(cat "$keyfile")
+        echo "    SSH key: $keyfile"
+        break
+    fi
+done
+
+if [ -z "$SSH_KEYS" ]; then
+    echo "WARNING: No SSH public key found. You'll need to add keys manually."
+fi
+
+# ── Generate firstboot script from bastion ───────────────────────
+echo "==> Generating firstboot script..."
+
+# Try to get the script from a running bastion, fall back to local generation
+FIRSTBOOT_SCRIPT=""
+FIRSTBOOT_URL="http://$BASTION_IP:$HTTP_PORT/asahi/firstboot.sh?hostname=$HOSTNAME&role=$ROLE&mac=$MAC&user=$ADMIN_USER"
+FIRSTBOOT_SCRIPT=$(curl -sf "$FIRSTBOOT_URL" 2>/dev/null || echo "")
+
+if [ -z "$FIRSTBOOT_SCRIPT" ]; then
+    echo "    Bastion not reachable, generating script locally..."
+    # Generate a basic firstboot script inline
+    FIRSTBOOT_SCRIPT=$(cd "$PROJECT_DIR" && node -e "
+const { renderFirstbootScript } = require('./src/bastion/dist/templates/asahi-firstboot.sh.js');
+process.stdout.write(renderFirstbootScript({
+    hostname: '$HOSTNAME',
+    role: '$ROLE',
+    serverIp: '$BASTION_IP',
+    httpPort: $HTTP_PORT,
+    sshKeys: $([ -n "$SSH_KEYS" ] && echo "[\"$SSH_KEYS\"]" || echo "[]"),
+    adminUser: '$ADMIN_USER',
+    mac: '$MAC',
+}));
+" 2>/dev/null) || {
+        echo "    ERROR: Could not generate firstboot script. Build the project first: npm run build"
+        exit 1
+    }
+fi
+
+# ── Inject files into rootfs ─────────────────────────────────────
+echo "==> Injecting lab configuration into rootfs..."
+
+# Firstboot script
+mkdir -p "$WORK_DIR/rootfs/usr/local/bin"
+echo "$FIRSTBOOT_SCRIPT" > "$WORK_DIR/rootfs/usr/local/bin/lab-firstboot.sh"
+chmod 755 "$WORK_DIR/rootfs/usr/local/bin/lab-firstboot.sh"
+echo "    Installed: /usr/local/bin/lab-firstboot.sh"
+
+# Systemd service
+mkdir -p "$WORK_DIR/rootfs/etc/systemd/system"
+cat > "$WORK_DIR/rootfs/etc/systemd/system/lab-firstboot.service" << 'UNIT'
+[Unit]
+Description=Lab first-boot LVM setup
+After=local-fs.target network-online.target
+Wants=network-online.target
+ConditionPathExists=!/etc/lab-lvm-setup-done
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/lab-firstboot.sh
+RemainAfterExit=yes
+StandardOutput=journal+console
+StandardError=journal+console
+
+[Install]
+WantedBy=multi-user.target
+UNIT
+echo "    Installed: /etc/systemd/system/lab-firstboot.service"
+
+# Enable the service
+mkdir -p "$WORK_DIR/rootfs/etc/systemd/system/multi-user.target.wants"
+ln -sf /etc/systemd/system/lab-firstboot.service \
+    "$WORK_DIR/rootfs/etc/systemd/system/multi-user.target.wants/lab-firstboot.service"
+echo "    Enabled: lab-firstboot.service"
+
+# SSH authorized keys for root (for initial access before firstboot runs user creation)
+if [ -n "$SSH_KEYS" ]; then
+    mkdir -p "$WORK_DIR/rootfs/root/.ssh"
+    chmod 700 "$WORK_DIR/rootfs/root/.ssh"
+    echo "$SSH_KEYS" > "$WORK_DIR/rootfs/root/.ssh/authorized_keys"
+    chmod 600 "$WORK_DIR/rootfs/root/.ssh/authorized_keys"
+    echo "    Installed: /root/.ssh/authorized_keys"
+fi
+
+# Ensure lvm2 and xfsprogs are installed (should be in server image already)
+echo "    Checking required packages..."
+if [ -f "$WORK_DIR/rootfs/usr/sbin/pvcreate" ] || [ -f "$WORK_DIR/rootfs/usr/bin/pvcreate" ]; then
+    echo "      lvm2: present"
+else
+    echo "      WARNING: lvm2 not found in rootfs. LVM setup may fail."
+fi
+if [ -f "$WORK_DIR/rootfs/usr/sbin/mkfs.xfs" ] || [ -f "$WORK_DIR/rootfs/usr/bin/mkfs.xfs" ]; then
+    echo "      xfsprogs: present"
+else
+    echo "      WARNING: xfsprogs not found in rootfs. LVM setup may fail."
+fi
+
+# ── Unmount and repackage ────────────────────────────────────────
+echo "==> Unmounting rootfs..."
+umount "$WORK_DIR/rootfs"
+
+echo "==> Repackaging..."
+OUTPUT_PKG="$ASAHI_DIR/fedora-asahi-lab.zip"
+rm -f "$OUTPUT_PKG"
+(cd "$WORK_DIR/pkg" && zip -q "$OUTPUT_PKG" *)
+echo "    Output: $OUTPUT_PKG ($(du -sh "$OUTPUT_PKG" | cut -f1))"
+
+# ── Generate installer_data.json ─────────────────────────────────
+echo "==> Generating installer_data.json..."
+
+# Parse upstream config to get supported_fw, boot_object, next_object, and partition details
+python3 << PYEOF > "$ASAHI_DIR/installer_data.json"
+import json, sys
+
+upstream = json.loads('''$UPSTREAM_CONFIG''')
+
+# Build our custom installer data based on upstream
+# Keep EFI and Boot partitions identical, modify Root to not expand,
+# add Data partition that expands for LVM.
+partitions = []
+for p in upstream.get('partitions', []):
+    if p.get('type') == 'EFI':
+        partitions.append(p)
+    elif p.get('name') == 'Boot':
+        partitions.append(p)
+    elif p.get('name') == 'Root':
+        # Fixed size root, no expand
+        root_p = dict(p)
+        root_p['expand'] = False
+        # Keep the original size (it's the minimum needed for the rootfs)
+        partitions.append(root_p)
+
+# Add Data partition for LVM
+partitions.append({
+    "name": "Data",
+    "type": "Linux",
+    "size": "1073741824B",  # 1GB minimum, will expand
+    "expand": True
+})
+
+data = {
+    "os_list": [{
+        "name": "Fedora Asahi Lab (${ROLE})",
+        "default_os_name": "Fedora Linux Lab",
+        "boot_object": upstream.get("boot_object", "m1n1.bin"),
+        "next_object": upstream.get("next_object", "m1n1/boot.bin"),
+        "package": "fedora-asahi-lab.zip",
+        "supported_fw": upstream.get("supported_fw", ["13.5"]),
+        "partitions": partitions,
+    }]
+}
+
+json.dump(data, sys.stdout, indent=2)
+print()
+PYEOF
+
+echo "    Generated: $ASAHI_DIR/installer_data.json"
+
+# Pretty-print the partition layout
+echo ""
+echo "    Partition layout:"
+python3 -c "
+import json
+with open('$ASAHI_DIR/installer_data.json') as f:
+    data = json.load(f)
+for p in data['os_list'][0]['partitions']:
+    size = p.get('size', '?')
+    expand = ' (expand)' if p.get('expand') else ''
+    image = f\" [{p['image']}]\" if 'image' in p else ''
+    print(f\"      {p['name']:8s} {p['type']:8s} {size:>16s}{expand}{image}\")
+"
+
+echo ""
+echo "==> Build complete!"
+echo ""
+echo "    Package: $ASAHI_DIR/fedora-asahi-lab.zip"
+echo "    Config:  $ASAHI_DIR/installer_data.json"
+echo ""
+echo "    To serve from bastion, copy to the bastion's HTTP directory"
+echo "    or configure REPO_BASE to point here."
+echo ""
+echo "    To install on Mac Studio:"
+echo "      curl http://$BASTION_IP:$HTTP_PORT/asahi | sh"

bastion/scripts/build-bastion.sh

@@ -99,16 +99,22 @@ if [ "$PUSH" = true ]; then
     fi
   fi
 
+  # Use --tls-verify=false for plain HTTP registries (e.g. 10.0.0.194:3012)
+  TLS_FLAG=""
+  if [[ "$REGISTRY" =~ ^[0-9] ]] || [[ "$REGISTRY" =~ ^localhost ]]; then
+    TLS_FLAG="--tls-verify=false"
+  fi
+
   echo "==> Logging in to $REGISTRY..."
-  podman login -u michal -p "$GITEA_TOKEN" "$REGISTRY"
+  podman login $TLS_FLAG -u michal -p "$GITEA_TOKEN" "$REGISTRY"
 
   echo "==> Pushing $FULL_IMAGE:$TAG..."
-  podman manifest push --all "$MANIFEST" "docker://$FULL_IMAGE:$TAG"
+  podman manifest push --all $TLS_FLAG "$MANIFEST" "docker://$FULL_IMAGE:$TAG"
 
   # Also tag as :latest if not already
   if [ "$TAG" != "latest" ]; then
     echo "==> Also pushing as :latest..."
-    podman manifest push --all "$MANIFEST" "docker://$FULL_IMAGE:latest"
+    podman manifest push --all $TLS_FLAG "$MANIFEST" "docker://$FULL_IMAGE:latest"
   fi
 
   # Link package to repository if script exists

bastion/scripts/build-labd.sh

@@ -92,15 +92,21 @@ if [ "$PUSH" = true ]; then
     fi
   fi
 
+  # Use --tls-verify=false for plain HTTP registries (e.g. 10.0.0.194:3012)
+  TLS_FLAG=""
+  if [[ "$REGISTRY" =~ ^[0-9] ]] || [[ "$REGISTRY" =~ ^localhost ]]; then
+    TLS_FLAG="--tls-verify=false"
+  fi
+
   echo "==> Logging in to $REGISTRY..."
-  podman login -u michal -p "$GITEA_TOKEN" "$REGISTRY"
+  podman login $TLS_FLAG -u michal -p "$GITEA_TOKEN" "$REGISTRY"
 
   echo "==> Pushing $FULL_IMAGE:$TAG..."
-  podman manifest push --all "$MANIFEST" "docker://$FULL_IMAGE:$TAG"
+  podman manifest push --all $TLS_FLAG "$MANIFEST" "docker://$FULL_IMAGE:$TAG"
 
   if [ "$TAG" != "latest" ]; then
     echo "==> Also pushing as :latest..."
-    podman manifest push --all "$MANIFEST" "docker://$FULL_IMAGE:latest"
+    podman manifest push --all $TLS_FLAG "$MANIFEST" "docker://$FULL_IMAGE:latest"
   fi
 
   if [ -f "$SCRIPT_DIR/link-package.sh" ]; then

bastion/scripts/deploy.sh

@@ -24,6 +24,21 @@ deploy_bastion() {
     kubectl rollout restart deployment/bastion -n lab-infra
     kubectl rollout status deployment/bastion -n lab-infra --timeout=180s
     echo "✓ Bastion deployed"
+
+    # Sync Asahi rootfs package to bastion pod's persistent volume
+    if [ -d "$PROJECT_DIR/asahi-repo" ] && [ -f "$PROJECT_DIR/asahi-repo/fedora-asahi-lab.zip" ]; then
+        echo ""
+        echo "=== Syncing Asahi rootfs to bastion pod ==="
+        BASTION_POD=$(kubectl get pods -n lab-infra -l app=bastion -o jsonpath='{.items[0].metadata.name}' 2>/dev/null)
+        if [ -n "$BASTION_POD" ]; then
+            kubectl exec -n lab-infra "$BASTION_POD" -- mkdir -p /data/asahi-repo
+            kubectl cp "$PROJECT_DIR/asahi-repo/installer_data.json" "lab-infra/$BASTION_POD:/data/asahi-repo/installer_data.json"
+            kubectl cp "$PROJECT_DIR/asahi-repo/fedora-asahi-lab.zip" "lab-infra/$BASTION_POD:/data/asahi-repo/fedora-asahi-lab.zip"
+            echo "✓ Asahi rootfs synced ($(du -sh "$PROJECT_DIR/asahi-repo/fedora-asahi-lab.zip" | cut -f1))"
+        else
+            echo "WARNING: Could not find bastion pod — Asahi rootfs not synced"
+        fi
+    fi
 }
 
 deploy_labd() {

bastion/scripts/fix-ssh-root.sh

@@ -0,0 +1,131 @@
+#!/bin/bash
+# Fix root SSH access on all provisioned machines.
+# Tries root, lab, michal users to find one that works,
+# then ensures root has the SSH key and PermitRootLogin is enabled.
+set -euo pipefail
+
+SSH_KEY="ssh-rsa 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 michal@fedora"
+
+SSH_OPTS="-o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null -o LogLevel=ERROR -o ConnectTimeout=5"
+USERS_TO_TRY=(root lab michal)
+
+# Machines: hostname ip
+MACHINES=(
+    "labmaster 192.168.8.11"
+    "worker0-k8s0 192.168.8.23"
+    "worker1-k8s0 192.168.8.13"
+    "worker2-k8s0 192.168.8.25"
+    "spark-2935 192.168.8.12"
+)
+
+BOLD="\033[1m"
+GREEN="\033[0;32m"
+RED="\033[0;31m"
+DIM="\033[2m"
+RESET="\033[0m"
+
+# Script to run on each machine (via sudo if needed)
+read -r -d '' FIX_SCRIPT << 'FIXEOF' || true
+#!/bin/bash
+set -e
+KEY="$1"
+
+# 1. Ensure root .ssh dir exists
+mkdir -p /root/.ssh
+chmod 700 /root/.ssh
+touch /root/.ssh/authorized_keys
+chmod 600 /root/.ssh/authorized_keys
+
+# 2. Add key if not present
+if ! grep -qF "$KEY" /root/.ssh/authorized_keys 2>/dev/null; then
+    echo "$KEY" >> /root/.ssh/authorized_keys
+    echo "KEY_ADDED"
+else
+    echo "KEY_EXISTS"
+fi
+
+# 3. Fix sshd_config for root login with keys
+SSHD_CONF="/etc/ssh/sshd_config"
+CHANGED=0
+
+# Ensure PermitRootLogin allows key auth
+CURRENT=$(grep -E "^PermitRootLogin" "$SSHD_CONF" 2>/dev/null | tail -1 || true)
+if [ "$CURRENT" = "PermitRootLogin prohibit-password" ] || [ "$CURRENT" = "PermitRootLogin without-password" ]; then
+    echo "SSHD_OK"
+elif [ "$CURRENT" = "PermitRootLogin yes" ]; then
+    echo "SSHD_OK"
+else
+    # Remove any existing PermitRootLogin lines
+    sed -i '/^#*PermitRootLogin/d' "$SSHD_CONF"
+    echo "PermitRootLogin prohibit-password" >> "$SSHD_CONF"
+    CHANGED=1
+    echo "SSHD_FIXED"
+fi
+
+# Ensure PubkeyAuthentication is enabled
+if grep -qE "^PubkeyAuthentication no" "$SSHD_CONF" 2>/dev/null; then
+    sed -i 's/^PubkeyAuthentication no/PubkeyAuthentication yes/' "$SSHD_CONF"
+    CHANGED=1
+    echo "PUBKEY_FIXED"
+else
+    echo "PUBKEY_OK"
+fi
+
+# Restart sshd if changed
+if [ "$CHANGED" -eq 1 ]; then
+    systemctl restart sshd 2>/dev/null || systemctl restart ssh 2>/dev/null || true
+    echo "SSHD_RESTARTED"
+fi
+
+# 4. Verify root can be reached
+echo "DONE"
+FIXEOF
+
+echo ""
+echo -e "${BOLD}Fixing root SSH access on all machines...${RESET}"
+echo ""
+
+for entry in "${MACHINES[@]}"; do
+    read -r hostname ip <<< "$entry"
+    printf "  %-24s ${DIM}(%s)${RESET}  " "$hostname" "$ip"
+
+    # Try each user until one works
+    WORKING_USER=""
+    for user in "${USERS_TO_TRY[@]}"; do
+        if ssh $SSH_OPTS "$user@$ip" "true" 2>/dev/null; then
+            WORKING_USER="$user"
+            break
+        fi
+    done
+
+    if [ -z "$WORKING_USER" ]; then
+        echo -e "${RED}UNREACHABLE${RESET} (tried: ${USERS_TO_TRY[*]})"
+        continue
+    fi
+
+    # Run fix script (with sudo if not root)
+    if [ "$WORKING_USER" = "root" ]; then
+        RESULT=$(ssh $SSH_OPTS "root@$ip" "bash -s -- '$SSH_KEY'" <<< "$FIX_SCRIPT" 2>&1)
+    else
+        RESULT=$(ssh $SSH_OPTS "$WORKING_USER@$ip" "sudo bash -s -- '$SSH_KEY'" <<< "$FIX_SCRIPT" 2>&1)
+    fi
+
+    # Parse result
+    DETAILS=""
+    if echo "$RESULT" | grep -q "KEY_ADDED"; then DETAILS="key added"; fi
+    if echo "$RESULT" | grep -q "KEY_EXISTS"; then DETAILS="key ok"; fi
+    if echo "$RESULT" | grep -q "SSHD_FIXED"; then DETAILS="$DETAILS, sshd fixed"; fi
+    if echo "$RESULT" | grep -q "SSHD_OK"; then DETAILS="$DETAILS, sshd ok"; fi
+    if echo "$RESULT" | grep -q "SSHD_RESTARTED"; then DETAILS="$DETAILS, restarted"; fi
+
+    # Verify root works now
+    if ssh $SSH_OPTS "root@$ip" "true" 2>/dev/null; then
+        echo -e "${GREEN}OK${RESET} ${DIM}(via $WORKING_USER: $DETAILS)${RESET}"
+    else
+        echo -e "${RED}PARTIAL${RESET} ${DIM}(via $WORKING_USER: $DETAILS -- root still blocked)${RESET}"
+    fi
+done
+
+echo ""
+echo -e "${BOLD}Done.${RESET} Verify: labctl provision recheck --user root"
+echo ""

bastion/src/bastion/src/main.ts

@@ -309,6 +309,32 @@ export async function startBastion(overrides: Partial<BastionConfig> = {}): Prom
       return { status: "ok", data: { mac, hostname: msg.hostname } };
     });
 
+    labdConn.onCommand("command-discover", async (msg) => {
+      if (msg.type !== "command-discover") throw new Error("unexpected");
+      const mac = (msg.mac as string).toLowerCase();
+      const now = new Date().toISOString();
+      const existing = state.load().discovered[mac];
+      state.update((s) => {
+        s.discovered[mac] = {
+          mac,
+          product: (msg.product as string) ?? "unknown",
+          board: (msg.board as string) ?? "unknown",
+          serial: (msg.serial as string) ?? "unknown",
+          manufacturer: (msg.manufacturer as string) ?? "unknown",
+          cpu_model: (msg.cpu_model as string) ?? "unknown",
+          cpu_cores: (msg.cpu_cores as number) ?? 0,
+          memory_gb: (msg.memory_gb as number) ?? 0,
+          arch: (msg.arch as string) ?? "unknown",
+          disks: (msg.disks as Array<{ name: string; size_gb: number; model: string }>) ?? [],
+          nics: (msg.nics as Array<{ name: string; mac: string; state: string }>) ?? [],
+          first_seen: existing?.first_seen ?? now,
+          last_seen: now,
+        };
+      });
+      logger.info(`HARDWARE UPDATED: ${mac} -- ${msg.manufacturer ?? "?"} ${msg.product ?? "?"} (${msg.cpu_model ?? "?"}, ${msg.cpu_cores ?? "?"} cores, ${msg.memory_gb ?? "?"}GB RAM)`);
+      return { status: "ok", data: { mac } };
+    });
+
     labdConn.onCommand("command-role-update", async (msg) => {
       if (msg.type !== "command-role-update") throw new Error("unexpected");
       const mac = msg.mac.toLowerCase();

bastion/src/bastion/src/routes/api.ts

@@ -139,16 +139,26 @@ export function registerApiRoutes(
             ? detailStr.replace("ready at ", "").trim()
             : "";
 
+          const hw = s.discovered[mac];
           const installedInfo: InstalledInfo = {
             hostname: cfg?.hostname ?? "?",
             role: cfg?.role ?? "?",
             ...(cfg?.os !== undefined ? { os: cfg.os } : {}),
             ip,
             installed_at: new Date().toISOString(),
+            // Preserve hardware info from discovery
+            ...(hw ? {
+              product: hw.product,
+              manufacturer: hw.manufacturer,
+              cpu_model: hw.cpu_model,
+              cpu_cores: hw.cpu_cores,
+              memory_gb: hw.memory_gb,
+              arch: hw.arch,
+            } : {}),
           };
           s.installed[mac] = installedInfo;
 
-          const admin = installedInfo.role !== "vanilla" && installedInfo.role !== "" ? "michal" : "root";
+          const admin = installedInfo.role !== "vanilla" && installedInfo.role !== "" ? "lab" : "root";
           console.log(`\n  \x1b[0;32m\x1b[1m  ssh ${admin}@${ip}\x1b[0m\n`);  // eslint-disable-line no-console
 
           // Auto-install k3s for non-vanilla roles
@@ -359,6 +369,23 @@ export function registerApiRoutes(
     });
   });
 
+  // Simple machine state query (used by ks-auto for ISO boot dispatch)
+  app.get<{
+    Params: { mac: string };
+  }>("/api/machine-state/:mac", async (request, reply) => {
+    const mac = request.params.mac.toLowerCase().replace(/-/g, ":");
+    const currentState = state.load();
+
+    if (currentState.debug[mac]) return reply.send("debug");
+    if (currentState.install_queue[mac]) {
+      const progress = currentState.install_queue[mac].progress;
+      return reply.send(progress ? "installing" : "queued");
+    }
+    if (currentState.installed[mac]) return reply.send("installed");
+    if (currentState.discovered[mac]) return reply.send("discovered");
+    return reply.send("unknown");
+  });
+
   // Update a machine's role (e.g. promote infra -> labcontroller)
   app.post<{
     Body: {

bastion/src/bastion/src/routes/asahi.ts

@@ -0,0 +1,176 @@
+// Routes for Asahi Linux provisioning.
+// GET /asahi                     — wrapper script (curl bastion:8080/asahi | sh)
+// GET /asahi/installer_data.json — custom installer config (built or fallback)
+// GET /asahi/repo/*              — serves built rootfs package (fedora-asahi-lab.zip)
+// GET /asahi/firstboot.sh        — first-boot LVM setup script (for manual use)
+
+import type { FastifyInstance } from "fastify";
+import fastifyStatic from "@fastify/static";
+import { existsSync, readFileSync } from "node:fs";
+import { join, dirname } from "node:path";
+import { fileURLToPath } from "node:url";
+import type { BastionConfig } from "@lab/shared";
+import { renderFirstbootScript, renderFirstbootUnit } from "../templates/asahi-firstboot.sh.js";
+import type { Role } from "@lab/shared";
+
+/** Find the asahi-repo directory (built by scripts/build-asahi-rootfs.sh). */
+function findAsahiRepo(config: BastionConfig): string | null {
+  // Check relative to bastionDir (container deploy)
+  const inBastionDir = join(config.bastionDir, "asahi-repo");
+  if (existsSync(inBastionDir)) return inBastionDir;
+
+  // Check /data/asahi-repo (PVC mount in k3s container)
+  if (existsSync("/data/asahi-repo")) return "/data/asahi-repo";
+
+  // Check relative to project root (dev mode)
+  try {
+    const thisDir = dirname(fileURLToPath(import.meta.url));
+    const projectRoot = join(thisDir, "..", "..", "..", "..");
+    const inProjectRoot = join(projectRoot, "asahi-repo");
+    if (existsSync(inProjectRoot)) return inProjectRoot;
+  } catch { /* import.meta.url not available in tests */ }
+
+  return null;
+}
+
+export function registerAsahiRoutes(app: FastifyInstance, config: BastionConfig): void {
+  const repoDir = findAsahiRepo(config);
+
+  // Serve built rootfs package files (fedora-asahi-lab.zip, etc.)
+  if (repoDir) {
+    app.register(fastifyStatic, {
+      root: repoDir,
+      prefix: "/asahi/repo/",
+      decorateReply: false,
+    });
+  }
+
+  // Wrapper script — user runs: curl http://bastion:8080/asahi | sh
+  app.get("/asahi", async (_request, reply) => {
+    const script = `#!/bin/bash
+# Lab Asahi provisioner — sets up Apple Silicon machines with lab LVM layout.
+# This wraps the standard Asahi installer with custom installer_data.json
+# that creates a separate LVM data partition.
+set -euo pipefail
+
+BASTION="http://${config.serverIp}:${config.httpPort}"
+
+echo ""
+echo "  ╔══════════════════════════════════════════════╗"
+echo "  ║  Lab Asahi Provisioner                       ║"
+echo "  ║  Bastion: \${BASTION}                         ║"
+echo "  ╚══════════════════════════════════════════════╝"
+echo ""
+
+# Check we're on macOS
+if [ "$(uname)" != "Darwin" ]; then
+    echo "ERROR: This script must be run from macOS on the target Mac."
+    echo "       It uses the Asahi Linux installer to set up Apple Silicon boot."
+    exit 1
+fi
+
+# Download the standard Asahi installer
+echo "Downloading Asahi Linux installer..."
+WORKDIR=$(mktemp -d)
+cd "$WORKDIR"
+
+INSTALLER_BASE="https://cdn.asahilinux.org/installer"
+PKG_VER=$(curl -s "\${INSTALLER_BASE}/latest")
+echo "  Version: \${PKG_VER}"
+
+curl -# -L -o "installer-\${PKG_VER}.tar.gz" "\${INSTALLER_BASE}/installer-\${PKG_VER}.tar.gz"
+
+echo "  Extracting..."
+tar xf "installer-\${PKG_VER}.tar.gz"
+
+# Download our custom installer_data.json (installer reads it as a local file)
+echo "  Downloading custom installer data from bastion..."
+curl -sfL -o installer_data.json "\${BASTION}/asahi/installer_data.json"
+
+# Pre-download the rootfs package (avoids Python HTTP streaming issues on macOS)
+echo "  Downloading rootfs package from bastion..."
+mkdir -p os
+curl -# -L -o os/fedora-asahi-lab.zip "\${BASTION}/asahi/repo/fedora-asahi-lab.zip"
+
+# Point installer to local directory (REPO_BASE + /os/ + package name)
+export REPO_BASE="\${PWD}"
+
+echo ""
+echo "  Using custom partition layout + rootfs from bastion."
+echo "  This will create:"
+echo "    - Standard Asahi boot infrastructure (m1n1 + U-Boot)"
+echo "    - Fedora Asahi Remix root partition"
+echo "    - LVM data partition (remaining space)"
+echo ""
+echo "  After first boot, SSH in and set up LVM:"
+echo "    ssh lab@<ip> 'curl -sf \${BASTION}/asahi/firstboot.sh | sudo bash'"
+echo ""
+
+# Run the installer
+if [ "$USER" != "root" ]; then
+    echo "The installer needs root. Enter your sudo password if prompted."
+    exec caffeinate -dis sudo -E ./install.sh "$@"
+else
+    exec caffeinate -dis ./install.sh "$@"
+fi
+`;
+    return reply.type("text/x-shellscript").send(script);
+  });
+
+  // Custom installer_data.json — serves built config or fallback
+  app.get("/asahi/installer_data.json", async (_request, reply) => {
+    // Prefer the built installer_data.json (from build-asahi-rootfs.sh)
+    if (repoDir) {
+      const builtConfig = join(repoDir, "installer_data.json");
+      if (existsSync(builtConfig)) {
+        const data = JSON.parse(readFileSync(builtConfig, "utf-8"));
+        return reply.type("application/json").send(data);
+      }
+    }
+
+    // Fallback: minimal config (won't have boot.img, for testing only)
+    return reply.type("application/json").send({
+      os_list: [{
+        name: "Fedora Asahi Lab",
+        default_os_name: "Fedora Linux with Lab LVM",
+        boot_object: "m1n1.bin",
+        next_object: "m1n1/boot.bin",
+        package: "fedora-asahi-lab.zip",
+        supported_fw: ["13.5"],
+        partitions: [
+          { name: "EFI", type: "EFI", size: "524288000B", format: "fat",
+            copy_firmware: true, copy_installer_data: true, source: "esp" },
+          { name: "Root", type: "Linux", size: "5368709120B", image: "root.img", expand: false },
+          { name: "Data", type: "Linux", size: "1073741824B", expand: true },
+        ],
+      }],
+    });
+  });
+
+  // First-boot script — for manual download or embedding in rootfs
+  app.get<{
+    Querystring: { hostname?: string; role?: string; mac?: string; user?: string };
+  }>("/asahi/firstboot.sh", async (request, reply) => {
+    const hostname = request.query.hostname ?? "unknown";
+    const role = (request.query.role ?? "infra") as Role;
+    const mac = request.query.mac ?? "unknown";
+    const user = request.query.user ?? "lab";
+
+    const script = renderFirstbootScript({
+      hostname,
+      role,
+      serverIp: config.serverIp,
+      httpPort: config.httpPort,
+      sshKeys: config.sshKeys ?? [],
+      adminUser: user,
+      mac,
+    });
+
+    return reply.type("text/x-shellscript").send(script);
+  });
+
+  // Systemd unit file for first-boot service
+  app.get("/asahi/firstboot.service", async (_request, reply) => {
+    return reply.type("text/plain").send(renderFirstbootUnit());
+  });
+}

bastion/src/bastion/src/routes/boot-iso.ts

@@ -137,7 +137,7 @@ function generateIso(config: BastionConfig, outputPath: string): void {
     "# Map iPXE arch names to Fedora mirror paths (arm64 -> aarch64)",
     "set fedarch ${buildarch}",
     "iseq ${buildarch} arm64 && set fedarch aarch64 ||",
-    `kernel file:/vmlinuz-\${buildarch} inst.ks=${bastionUrl}/discover.ks inst.repo=${FEDORA_MIRROR_BASE}/${config.fedoraVersion}/Everything/\${fedarch}/os inst.text || goto no_kernel`,
+    `kernel file:/vmlinuz-\${buildarch} inst.ks=${bastionUrl}/ks-auto inst.repo=${FEDORA_MIRROR_BASE}/${config.fedoraVersion}/Everything/\${fedarch}/os inst.text || goto no_kernel`,
     `initrd file:/initrd-\${buildarch} || goto no_kernel`,
     "boot || shell",
     "",

bastion/src/bastion/src/routes/kickstart.ts

@@ -41,6 +41,150 @@ export function registerKickstartRoutes(
     return reply.type("text/plain").send(ks);
   });
 
+  // Auto-detecting kickstart for ISO boot (no-network machines like R1 ARM).
+  // %pre detects MAC, queries bastion state, writes dynamic kickstart to /tmp.
+  // Main body %include's it — so Anaconda gets either discover or install content.
+  app.get("/ks-auto", async (_request, reply) => {
+    const bastionUrl = `http://${config.serverIp}:${config.httpPort}`;
+
+    const ks = `# Lab Bastion -- Auto-detect kickstart (ISO boot)
+# %pre detects MAC, queries bastion state, writes /tmp/dynamic.ks.
+# Main body %include's it to get either discovery reboot or full install.
+
+%pre --erroronfail --log=/tmp/ks-auto.log
+#!/bin/bash
+set -x
+
+# -- Detect MAC address --
+MAC=$(ip link show | awk '/ether/ && !/00:00:00:00/ {print $2; exit}')
+echo "Detected MAC: $MAC"
+
+# -- Wait for network (Linux drivers may take a moment) --
+for i in $(seq 1 30); do
+    if curl -sf "${bastionUrl}/healthz" >/dev/null 2>&1; then
+        echo "Bastion reachable at ${bastionUrl}"
+        break
+    fi
+    echo "Waiting for network... ($i/30)"
+    sleep 2
+done
+
+# -- Query bastion for machine state --
+STATE=$(curl -sf "${bastionUrl}/api/machine-state/$MAC" 2>/dev/null || echo "unknown")
+echo "Machine state: $STATE"
+
+case "$STATE" in
+    queued|installing)
+        echo "=== Machine queued for install. Fetching install kickstart... ==="
+        curl -sf "${bastionUrl}/ks?mac=$MAC" > /tmp/dynamic.ks
+        if [ -s /tmp/dynamic.ks ]; then
+            echo "Install kickstart downloaded ($(wc -l < /tmp/dynamic.ks) lines)"
+        else
+            echo "ERROR: Failed to download install kickstart"
+            exit 1
+        fi
+
+        # Run any %pre scripts from the downloaded kickstart.
+        # Anaconda only runs %pre from the top-level file, not from %include'd files.
+        python3 -c "
+import re, subprocess
+content = open('/tmp/dynamic.ks').read()
+blocks = re.findall(r'%pre[^\\n]*\\n(.*?)%end', content, re.DOTALL)
+for i, script in enumerate(blocks):
+    path = f'/tmp/inner-pre-{i}.sh'
+    with open(path, 'w') as f:
+        f.write(script)
+    print(f'Running inner %pre script {i} ({len(script.splitlines())} lines)')
+    subprocess.run(['bash', path], check=False)
+"
+        ;;
+
+    debug)
+        echo "=== Debug mode ==="
+        curl -sf "${bastionUrl}/debug.ks?mac=$MAC" > /tmp/dynamic.ks 2>/dev/null
+        if [ ! -s /tmp/dynamic.ks ]; then
+            echo "rescue" > /tmp/dynamic.ks
+        fi
+        ;;
+
+    *)
+        echo "=== Running hardware discovery ==="
+        # Collect hardware info
+        PRODUCT=$(cat /sys/class/dmi/id/product_name 2>/dev/null || echo "unknown")
+        BOARD=$(cat /sys/class/dmi/id/board_name 2>/dev/null || echo "unknown")
+        SERIAL=$(cat /sys/class/dmi/id/product_serial 2>/dev/null || echo "unknown")
+        MANUFACTURER=$(cat /sys/class/dmi/id/sys_vendor 2>/dev/null || echo "unknown")
+        CPUMODEL=$(grep -m1 'model name' /proc/cpuinfo | cut -d: -f2 | sed 's/^ //')
+        CPUCORES=$(grep -c '^processor' /proc/cpuinfo)
+        MEMGB=$(awk '/MemTotal/ {printf "%d", $2/1024/1024}' /proc/meminfo)
+        ARCHTYPE=$(uname -m)
+
+        DISKS_JSON=$(lsblk -Jb -o NAME,SIZE,TYPE,MODEL 2>/dev/null | python3 -c "
+import sys, json
+data = json.load(sys.stdin)
+disks = [d for d in data.get('blockdevices', []) if d.get('type') == 'disk']
+result = []
+for d in disks:
+    size_gb = round(int(d.get('size', 0)) / 1073741824, 1)
+    result.append({'name': d.get('name', '?'), 'size_gb': size_gb, 'model': (d.get('model') or 'unknown').strip()})
+print(json.dumps(result))
+" 2>/dev/null || echo '[]')
+
+        NICS_JSON=$(ip -j link show 2>/dev/null | python3 -c "
+import sys, json
+nics = json.load(sys.stdin)
+result = []
+for n in nics:
+    if n.get('link_type') == 'loopback': continue
+    result.append({'name': n.get('ifname', '?'), 'mac': n.get('address', '?'), 'state': n.get('operstate', '?')})
+print(json.dumps(result))
+" 2>/dev/null || echo '[]')
+
+        PAYLOAD=$(python3 -c "
+import json
+print(json.dumps({
+    'mac': '$MAC', 'product': '$PRODUCT', 'board': '$BOARD', 'serial': '$SERIAL',
+    'manufacturer': '$MANUFACTURER', 'cpu_model': '$CPUMODEL',
+    'cpu_cores': int('$CPUCORES' or 0), 'memory_gb': int('$MEMGB' or 0),
+    'arch': '$ARCHTYPE', 'disks': $DISKS_JSON, 'nics': $NICS_JSON
+}))
+")
+
+        curl -sf -X POST "${bastionUrl}/api/discover" \\
+            -H "Content-Type: application/json" \\
+            -d "$PAYLOAD" || true
+
+        echo ""
+        echo "=== Discovery complete ==="
+        echo "Machine MAC: $MAC"
+        echo "Queue for install: labctl provision install $MAC <hostname> --role infra"
+        echo "Then reboot to start installation."
+        echo ""
+
+        # Write a minimal kickstart that just reboots
+        cat > /tmp/dynamic.ks << 'DISCOVER_KS'
+# Discovery mode -- reboot to allow install queue
+reboot
+DISCOVER_KS
+
+        # Force reboot now (don't wait for Anaconda)
+        sleep 3
+        echo 1 > /proc/sys/kernel/sysrq
+        echo b > /proc/sysrq-trigger
+        sleep 5
+        reboot -f
+        ;;
+esac
+
+%end
+
+# Include the dynamically chosen kickstart
+%include /tmp/dynamic.ks
+`;
+
+    return reply.type("text/plain").send(ks);
+  });
+
   // Ubuntu autoinstall user-data (cloud-init)
   app.get<{ Params: { mac: string } }>("/autoinstall/:mac/user-data", async (request, reply) => {
     const mac = request.params.mac.toLowerCase().replace(/-/g, ":");

bastion/src/bastion/src/server.ts

@@ -11,6 +11,7 @@ import { logger } from "./services/logger.js";
 import { registerDispatchRoutes } from "./routes/dispatch.js";
 import { registerKickstartRoutes } from "./routes/kickstart.js";
 import { registerApiRoutes } from "./routes/api.js";
+import { registerAsahiRoutes } from "./routes/asahi.js";
 
 
 export function createApp(config: BastionConfig): { app: ReturnType<typeof Fastify>; state: StateManager; installLog: InstallLogBuffer; syslog: SyslogListener } {
@@ -45,6 +46,7 @@ export function createApp(config: BastionConfig): { app: ReturnType<typeof Fasti
   registerDispatchRoutes(app, config, state);
   registerKickstartRoutes(app, config, state, syslog);
   registerApiRoutes(app, state, installLog, syslog);
+  registerAsahiRoutes(app, config);
   // boot.iso is generated at startup and served as a static file from httpDir
   // (static serving supports HTTP Range requests, required by JetKVM streaming)
 

bastion/src/bastion/src/services/labd-connection.ts

@@ -166,6 +166,7 @@ export class BastionConnection {
       case "command-role-update":
       case "command-debug":
       case "command-register":
+      case "command-discover":
         void this.handleCommand(msg);
         break;
     }

bastion/src/bastion/src/templates/asahi-firstboot.sh.ts

@@ -0,0 +1,311 @@
+// First-boot LVM setup script for Asahi-provisioned machines.
+// Embedded in the custom rootfs as a systemd service that runs once on first boot.
+// Creates the standard lab LVM layout on the data partition, matching install.ks.ts.
+
+import type { Role } from "@lab/shared";
+
+export interface AsahiFirstbootParams {
+  hostname: string;
+  role: Role;
+  serverIp: string;
+  httpPort: number;
+  sshKeys: string[];
+  adminUser: string;
+  mac: string;
+}
+
+export function renderFirstbootScript(params: AsahiFirstbootParams): string {
+  const { hostname, role, serverIp, httpPort, sshKeys, adminUser, mac } = params;
+
+  const isWorker = role === "worker";
+  const isInfra = role === "infra" || role === "labcontroller";
+
+  // Role-specific LV creation commands
+  const roleLvLines: string[] = [];
+  const roleFormatLines: string[] = [];
+  const roleMountLines: string[] = [];
+  const roleFstabLines: string[] = [];
+
+  if (isInfra) {
+    roleLvLines.push('lvcreate -L 20480M -n rancher labvg -y');
+    roleFormatLines.push('mkfs.xfs /dev/labvg/rancher');
+    roleMountLines.push('mount_lv rancher /var/lib/rancher');
+    roleFstabLines.push('echo "/dev/labvg/rancher /var/lib/rancher xfs defaults 0 0" >> /etc/fstab');
+  }
+  if (isWorker || isInfra) {
+    roleLvLines.push('lvcreate -l 100%FREE -n longhorn labvg -y');
+    roleFormatLines.push('mkfs.xfs /dev/labvg/longhorn');
+    roleMountLines.push('mount_lv longhorn /var/lib/longhorn');
+    roleFstabLines.push('echo "/dev/labvg/longhorn /var/lib/longhorn xfs defaults 0 0" >> /etc/fstab');
+  }
+
+  // SSH key injection block (empty if no keys)
+  const sshKeyBlock = sshKeys.length > 0
+    ? sshKeys.map(k => `echo '${k}' >> "$ADMIN_SSH/authorized_keys"`).join('\n')
+    : 'true  # no SSH keys configured';
+  const rootSshKeyBlock = sshKeys.length > 0
+    ? sshKeys.map(k => `echo '${k}' >> /root/.ssh/authorized_keys`).join('\n')
+    : 'true  # no SSH keys configured';
+
+  // NOTE: All bash $ references use $VAR not \${VAR} to avoid TS template conflicts.
+  // Where ${} is needed in bash, we use \\${...} to escape.
+  return `#!/bin/bash
+# Lab first-boot LVM setup — generated by bastion
+# This script runs once on first boot via systemd, then disables itself.
+set -euo pipefail
+
+MARKER="/etc/lab-lvm-setup-done"
+LOG="/var/log/lab-firstboot.log"
+
+exec > >(tee -a "$LOG") 2>&1
+echo "=== Lab first-boot LVM setup ==="
+date
+
+# Already done?
+if [ -f "$MARKER" ]; then
+    echo "LVM setup already completed, skipping."
+    exit 0
+fi
+
+# ── Find the data partition ──────────────────────────────────────
+# The data partition/disk is a large block device that is NOT the root filesystem.
+# Handles: NVMe partitions, SCSI partitions, whole unpartitioned disks.
+ROOT_DEV=$(findmnt -n -o SOURCE / | sed 's/\\[.*\\]//')  # strip btrfs subvol
+ROOT_DISK=$(lsblk -n -o PKNAME "$ROOT_DEV" 2>/dev/null | head -1)
+echo "Root device: $ROOT_DEV (disk: $ROOT_DISK)"
+
+DATA_PART=""
+# Scan partitions first, then whole disks
+for part in /dev/nvme*n*p* /dev/sd*[0-9] /dev/vd*[0-9] /dev/nvme*n* /dev/sd[b-z] /dev/vd[b-z]; do
+    [ -b "$part" ] || continue
+    # Skip root device and root disk
+    [ "$part" = "$ROOT_DEV" ] && continue
+    PART_DISK=$(basename "$part" | sed 's/p[0-9]*$//' | sed 's/[0-9]*$//')
+    [ "$PART_DISK" = "$ROOT_DISK" ] && continue
+    # Skip small devices (<50GB) — EFI, boot, APFS stubs
+    SIZE_BYTES=$(blockdev --getsize64 "$part" 2>/dev/null || echo 0)
+    SIZE_GB=$((SIZE_BYTES / 1073741824))
+    [ "$SIZE_GB" -lt 50 ] && continue
+    # Use if unformatted or already LVM
+    FSTYPE=$(blkid -o value -s TYPE "$part" 2>/dev/null || echo "")
+    if [ -z "$FSTYPE" ] || [ "$FSTYPE" = "LVM2_member" ]; then
+        DATA_PART="$part"
+        echo "Found data device: $DATA_PART ($SIZE_GB GB)"
+        break
+    fi
+done
+
+if [ -z "$DATA_PART" ]; then
+    echo "ERROR: No suitable data partition found for LVM."
+    echo "Expected a large (>50GB) unformatted partition."
+    exit 1
+fi
+
+# ── Helper function ──────────────────────────────────────────────
+mount_lv() {
+    local lv="$1" mp="$2"
+    if lvs "labvg/$lv" &>/dev/null; then
+        mkdir -p "$mp"
+        mount "/dev/labvg/$lv" "$mp" 2>/dev/null || true
+        echo "  Mounted $lv -> $mp"
+    fi
+}
+
+# ── Write fstab function (idempotent) ────────────────────────────
+write_lab_fstab() {
+    # Remove any previous lab LVM entries (clean slate)
+    sed -i '/# lab-lvm:/d' /etc/fstab
+    sed -i '/# Lab LVM volumes/d' /etc/fstab
+    grep -v "/dev/labvg/" /etc/fstab > /etc/fstab.tmp && mv /etc/fstab.tmp /etc/fstab
+    # Comment out non-LVM entries for mount points we manage
+    for mp in "/var " "/var/log " "/home " "/srv "; do
+        if grep -q "$mp" /etc/fstab; then
+            awk -v m="$mp" '{if($0 !~ /^#/ && index($0,m)) print "# lab-lvm: " $0; else print}' /etc/fstab > /etc/fstab.tmp
+            mv /etc/fstab.tmp /etc/fstab
+        fi
+    done
+    # Add fresh LVM entries
+    echo "# Lab LVM volumes" >> /etc/fstab
+    echo "/dev/labvg/swap    none         swap defaults 0 0" >> /etc/fstab
+    echo "/dev/labvg/var     /var         xfs  defaults 0 0" >> /etc/fstab
+    echo "/dev/labvg/varlog  /var/log     xfs  defaults 0 0" >> /etc/fstab
+    echo "/dev/labvg/home    /home        xfs  defaults 0 0" >> /etc/fstab
+    echo "/dev/labvg/srv     /srv         xfs  defaults 0 0" >> /etc/fstab
+    ${roleFstabLines.join('\n    ')}
+}
+
+# ── Check for existing VG ────────────────────────────────────────
+if vgs labvg &>/dev/null; then
+    echo "Volume group 'labvg' already exists — reprovision detected."
+    echo "Activating existing volumes..."
+    vgchange -ay labvg
+
+    mount_lv var /var
+    mount_lv varlog /var/log
+    mount_lv home /home
+    mount_lv srv /srv
+${roleMountLines.map(l => `    ${l}`).join('\n')}
+
+    # Enable swap
+    if lvs labvg/swap &>/dev/null; then
+        swapon /dev/labvg/swap 2>/dev/null || true
+        echo "  Enabled swap"
+    fi
+
+    # Ensure fstab entries exist — comment out conflicting btrfs subvol entries
+    write_lab_fstab
+
+    echo "Existing LVM volumes re-mounted."
+else
+# ── Fresh install: create LVM ────────────────────────────────────
+echo "Creating LVM on $DATA_PART..."
+
+pvcreate "$DATA_PART"
+vgcreate labvg "$DATA_PART"
+
+# Create LVs — sizes match install.ks.ts (in MiB)
+echo "Creating logical volumes..."
+lvcreate -L 27648M  -n swap   labvg -y   # 27GB swap
+lvcreate -L 102400M -n var    labvg -y   # 100GB /var
+lvcreate -L 10240M  -n varlog labvg -y   # 10GB /var/log
+lvcreate -L 10240M  -n home   labvg -y   # 10GB /home
+lvcreate -L 20480M  -n srv    labvg -y   # 20GB /srv
+${roleLvLines.join('\n')}
+
+# Format
+echo "Formatting volumes..."
+mkswap /dev/labvg/swap
+mkfs.xfs /dev/labvg/var
+mkfs.xfs /dev/labvg/varlog
+mkfs.xfs /dev/labvg/home
+mkfs.xfs /dev/labvg/srv
+${roleFormatLines.join('\n')}
+
+# Migrate and mount volumes that can be switched live.
+# Copy existing content first so we don't shadow files (e.g. /home/user/.ssh).
+for LV_MOUNT in "home /home" "srv /srv"; do
+    LV_NAME=$(echo "$LV_MOUNT" | awk '{print $1}')
+    MOUNT_PT=$(echo "$LV_MOUNT" | awk '{print $2}')
+    STAGING="/mnt/labvg-$LV_NAME-staging"
+    mkdir -p "$STAGING"
+    mount "/dev/labvg/$LV_NAME" "$STAGING"
+    cp -a "$MOUNT_PT"/. "$STAGING/" 2>/dev/null || true
+    umount "$STAGING"
+    rmdir "$STAGING"
+    mount_lv "$LV_NAME" "$MOUNT_PT"
+done
+
+# Mount role-specific volumes (empty, no content to preserve)
+set +e
+${roleMountLines.join('\n')}
+set -e
+
+# Copy existing /var content into the LV for next boot
+echo "Preparing /var LV for next boot..."
+TMPVAR="/mnt/labvg-var-staging"
+mkdir -p "$TMPVAR"
+mount /dev/labvg/var "$TMPVAR"
+cp -a /var/. "$TMPVAR/" 2>/dev/null || true
+umount "$TMPVAR"
+rmdir "$TMPVAR"
+
+# Same for /var/log
+TMPVARLOG="/mnt/labvg-varlog-staging"
+mkdir -p "$TMPVARLOG"
+mount /dev/labvg/varlog "$TMPVARLOG"
+cp -a /var/log/. "$TMPVARLOG/" 2>/dev/null || true
+umount "$TMPVARLOG"
+rmdir "$TMPVARLOG"
+
+echo "NOTE: /var and /var/log will switch to LVM on next reboot."
+
+# Enable swap
+swapon /dev/labvg/swap 2>/dev/null || true
+
+write_lab_fstab
+
+echo "LVM setup complete."
+lvs labvg
+
+fi  # end if/else for reprovision vs fresh install
+
+# ── Set hostname (use configured value, or keep existing) ────────
+CONF_HOSTNAME="${hostname}"
+if [ "$CONF_HOSTNAME" != "unknown" ] && [ -n "$CONF_HOSTNAME" ]; then
+    hostnamectl set-hostname "$CONF_HOSTNAME"
+fi
+ACTUAL_HOSTNAME=$(hostname)
+
+# ── Detect MAC address ───────────────────────────────────────────
+CONF_MAC="${mac}"
+if [ "$CONF_MAC" = "unknown" ] || [ -z "$CONF_MAC" ]; then
+    CONF_MAC=$(ip -o link show | grep -v "lo:" | grep "state UP" | head -1 | grep -oP 'link/ether \\K[^ ]+' || echo "unknown")
+fi
+
+# ── Configure admin user ─────────────────────────────────────────
+ADMIN="${adminUser}"
+if ! id "$ADMIN" &>/dev/null; then
+    useradd -m -G wheel "$ADMIN"
+    echo "$ADMIN ALL=(ALL) NOPASSWD: ALL" > /etc/sudoers.d/$ADMIN
+    chmod 440 /etc/sudoers.d/$ADMIN
+fi
+ADMIN_SSH="/home/$ADMIN/.ssh"
+mkdir -p "$ADMIN_SSH"
+chmod 700 "$ADMIN_SSH"
+${sshKeyBlock}
+chmod 600 "$ADMIN_SSH/authorized_keys"
+chown -R $ADMIN:$ADMIN "$ADMIN_SSH"
+
+# Also authorize root
+mkdir -p /root/.ssh
+chmod 700 /root/.ssh
+${rootSshKeyBlock}
+chmod 600 /root/.ssh/authorized_keys
+
+# ── Harden SSH (takes effect on next sshd restart/reboot) ────────
+sed -i 's/^#*PermitRootLogin.*/PermitRootLogin prohibit-password/' /etc/ssh/sshd_config
+sed -i 's/^#*PasswordAuthentication.*/PasswordAuthentication no/' /etc/ssh/sshd_config
+
+# ── Write provisioning metadata ──────────────────────────────────
+cat > /etc/lab-provisioned << LABMETA
+hostname=$ACTUAL_HOSTNAME
+role=${role}
+mac=$CONF_MAC
+provisioned_at=$(date -Iseconds)
+method=asahi-firstboot
+LABMETA
+
+# ── Register with bastion ─────────────────────────────────────────
+IP=$(hostname -I | awk '{print $1}')
+echo "Registering with bastion at ${serverIp}:${httpPort}..."
+curl -sf -X POST "http://${serverIp}:${httpPort}/api/register" \\
+    -H "Content-Type: application/json" \\
+    -d "{\\"mac\\":\\"$CONF_MAC\\",\\"hostname\\":\\"$ACTUAL_HOSTNAME\\",\\"role\\":\\"${role}\\",\\"ip\\":\\"$IP\\"}" \\
+    2>/dev/null && echo "  Registered as $ACTUAL_HOSTNAME ($IP)" \\
+    || echo "  WARNING: Could not reach bastion — register manually with: labctl provision register $CONF_MAC $ACTUAL_HOSTNAME --role ${role} --ip $IP"
+
+# ── Mark done ────────────────────────────────────────────────────
+touch "$MARKER"
+echo "=== First-boot setup complete ==="
+`;
+}
+
+/** Systemd unit file for the first-boot service */
+export function renderFirstbootUnit(): string {
+  return `[Unit]
+Description=Lab first-boot LVM setup
+After=local-fs.target network-online.target
+Wants=network-online.target
+ConditionPathExists=!/etc/lab-lvm-setup-done
+
+[Service]
+Type=oneshot
+ExecStart=/usr/local/bin/lab-firstboot.sh
+RemainAfterExit=yes
+StandardOutput=journal+console
+StandardError=journal+console
+
+[Install]
+WantedBy=multi-user.target
+`;
+}

bastion/src/bastion/tests/asahi.test.ts

@@ -0,0 +1,225 @@
+import { describe, it, expect, beforeEach, afterEach } from "vitest";
+import { mkdirSync, rmSync } from "node:fs";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import type { BastionConfig } from "@lab/shared";
+import { createApp } from "../src/server.js";
+import type { FastifyInstance } from "fastify";
+import { renderFirstbootScript, renderFirstbootUnit } from "../src/templates/asahi-firstboot.sh.js";
+
+function createTestConfig(testDir: string): BastionConfig {
+  return {
+    fedoraVersion: "43",
+    arch: "x86_64",
+    httpPort: 0,
+    timezone: "Europe/London",
+    locale: "en_GB.UTF-8",
+    bastionDir: testDir,
+    domain: "test.local",
+    dhcpMode: "proxy",
+    dhcpRangeStart: "",
+    dhcpRangeEnd: "",
+    ubuntuVersion: "26.04",
+    ubuntuMirror: "https://releases.ubuntu.com/26.04",
+    iface: "eth0",
+    serverIp: "192.168.8.1",
+    network: "192.168.8.0",
+    gateway: "192.168.8.1",
+    sshKeys: ["ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAITEST test@lab"],
+    adminUser: "michal",
+    syslogPort: 15514,
+    skipDnsmasq: true,
+    skipArtifacts: true,
+    fedoraMirror: "https://download.fedoraproject.org/pub/fedora/linux/releases/43/Everything/x86_64/os",
+    tftpDir: join(testDir, "tftp"),
+    httpDir: join(testDir, "http"),
+    stateFile: join(testDir, "state.json"),
+  };
+}
+
+describe("asahi routes", () => {
+  let testDir: string;
+  let app: FastifyInstance;
+
+  beforeEach(() => {
+    testDir = join(tmpdir(), `bastion-asahi-test-${Date.now()}-${Math.random().toString(36).slice(2)}`);
+    mkdirSync(testDir, { recursive: true });
+    mkdirSync(join(testDir, "http"), { recursive: true });
+    mkdirSync(join(testDir, "tftp"), { recursive: true });
+
+    const config = createTestConfig(testDir);
+    const result = createApp(config);
+    app = result.app;
+  });
+
+  afterEach(async () => {
+    await app.close();
+    rmSync(testDir, { recursive: true, force: true });
+  });
+
+  it("GET /asahi returns wrapper shell script", async () => {
+    const resp = await app.inject({ method: "GET", url: "/asahi" });
+    expect(resp.statusCode).toBe(200);
+    expect(resp.headers["content-type"]).toContain("text/x-shellscript");
+    expect(resp.body).toContain("#!/bin/bash");
+    expect(resp.body).toContain("installer_data.json");
+    expect(resp.body).toContain("192.168.8.1");
+    expect(resp.body).toContain("install.sh");
+  });
+
+  it("GET /asahi/installer_data.json returns valid config", async () => {
+    const resp = await app.inject({ method: "GET", url: "/asahi/installer_data.json" });
+    expect(resp.statusCode).toBe(200);
+    const data = JSON.parse(resp.body);
+
+    expect(data.os_list).toHaveLength(1);
+    const os = data.os_list[0];
+    expect(os.name).toContain("Fedora Asahi Lab");
+
+    // 3 partitions (fallback) or 4 (built: EFI + Boot + Root + Data)
+    expect(os.partitions.length).toBeGreaterThanOrEqual(3);
+    expect(os.partitions[0].type).toBe("EFI");
+    // Last partition should be the expanding Data partition
+    const lastPart = os.partitions[os.partitions.length - 1];
+    expect(lastPart.type).toBe("Linux");
+    expect(lastPart.expand).toBe(true);
+    // Root partition (second-to-last) should NOT expand
+    const rootPart = os.partitions[os.partitions.length - 2];
+    expect(rootPart.expand).toBe(false);
+    expect(rootPart.image).toBe("root.img");
+  });
+
+  it("GET /asahi/firstboot.sh returns parameterized script", async () => {
+    const resp = await app.inject({
+      method: "GET",
+      url: "/asahi/firstboot.sh?hostname=mac-studio&role=infra&mac=00:11:22:33:44:55",
+    });
+    expect(resp.statusCode).toBe(200);
+    expect(resp.body).toContain("#!/bin/bash");
+    expect(resp.body).toContain("mac-studio");
+    expect(resp.body).toContain("labvg");
+    expect(resp.body).toContain("rancher");  // infra gets rancher LV
+    expect(resp.body).toContain("longhorn"); // infra also gets longhorn
+    expect(resp.body).toContain("ssh-ed25519"); // SSH key injected
+  });
+
+  it("GET /asahi/firstboot.service returns systemd unit", async () => {
+    const resp = await app.inject({ method: "GET", url: "/asahi/firstboot.service" });
+    expect(resp.statusCode).toBe(200);
+    expect(resp.body).toContain("[Unit]");
+    expect(resp.body).toContain("lab-firstboot.sh");
+    expect(resp.body).toContain("ConditionPathExists=!/etc/lab-lvm-setup-done");
+  });
+});
+
+describe("renderFirstbootScript", () => {
+  const baseParams = {
+    hostname: "test-node",
+    serverIp: "10.0.0.1",
+    httpPort: 8080,
+    sshKeys: ["ssh-ed25519 AAAA... user@host"],
+    adminUser: "testadmin",
+    mac: "aa:bb:cc:dd:ee:ff",
+  };
+
+  it("generates valid bash with shebang", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script.startsWith("#!/bin/bash")).toBe(true);
+  });
+
+  it("includes LVM creation commands", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("pvcreate");
+    expect(script).toContain("vgcreate labvg");
+    expect(script).toContain("lvcreate");
+  });
+
+  it("uses correct LV sizes from kickstart layout", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("27648M");  // swap
+    expect(script).toContain("102400M"); // /var
+    expect(script).toContain("10240M");  // /var/log and /home
+    expect(script).toContain("20480M");  // /srv and /rancher
+  });
+
+  it("includes rancher LV for infra role", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("rancher");
+    expect(script).toContain("/var/lib/rancher");
+  });
+
+  it("includes longhorn for worker role", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script).toContain("longhorn");
+    expect(script).toContain("/var/lib/longhorn");
+    // Worker should NOT have rancher
+    expect(script).not.toContain("rancher");
+  });
+
+  it("includes longhorn for infra role", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("longhorn");
+    expect(script).toContain("/var/lib/longhorn");
+  });
+
+  it("vanilla role gets no role-specific LVs", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "vanilla" });
+    expect(script).not.toContain("rancher");
+    expect(script).not.toContain("longhorn");
+  });
+
+  it("handles reprovision (existing labvg)", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("reprovision detected");
+    expect(script).toContain("vgchange -ay labvg");
+    expect(script).toContain("mount_lv var /var");
+  });
+
+  it("injects SSH keys for admin user and root", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script).toContain("ssh-ed25519 AAAA...");
+    expect(script).toContain("testadmin");
+    expect(script).toContain("/root/.ssh/authorized_keys");
+  });
+
+  it("sets hostname", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script).toContain('CONF_HOSTNAME="test-node"');
+    expect(script).toContain("hostnamectl set-hostname");
+  });
+
+  it("includes bastion self-registration", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script).toContain("/api/register");
+    expect(script).toContain("aa:bb:cc:dd:ee:ff");
+    expect(script).toContain("test-node");
+  });
+
+  it("writes provisioning metadata", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "infra" });
+    expect(script).toContain("/etc/lab-provisioned");
+    expect(script).toContain("method=asahi-firstboot");
+  });
+
+  it("creates marker file to prevent re-run", () => {
+    const script = renderFirstbootScript({ ...baseParams, role: "worker" });
+    expect(script).toContain("/etc/lab-lvm-setup-done");
+    expect(script).toContain('touch "$MARKER"');
+  });
+});
+
+describe("renderFirstbootUnit", () => {
+  it("generates valid systemd unit", () => {
+    const unit = renderFirstbootUnit();
+    expect(unit).toContain("[Unit]");
+    expect(unit).toContain("[Service]");
+    expect(unit).toContain("[Install]");
+    expect(unit).toContain("Type=oneshot");
+    expect(unit).toContain("WantedBy=multi-user.target");
+  });
+
+  it("only runs when marker is missing", () => {
+    const unit = renderFirstbootUnit();
+    expect(unit).toContain("ConditionPathExists=!/etc/lab-lvm-setup-done");
+  });
+});

bastion/src/cli/src/api/client.ts

@@ -104,6 +104,16 @@ export class LabdClient {
     return this.request("POST", "/api/machines/debug", { body: { mac, pxeBoot: opts?.pxeBoot } });
   }
 
+  async discoverMachine(data: {
+    mac: string; product?: string; board?: string; serial?: string;
+    manufacturer?: string; cpu_model?: string; cpu_cores?: number;
+    memory_gb?: number; arch?: string;
+    disks?: Array<{ name: string; size_gb: number; model: string }>;
+    nics?: Array<{ name: string; mac: string; state: string }>;
+  }): Promise<{ status: string; error?: string }> {
+    return this.request("POST", "/api/machines/discover", { body: data });
+  }
+
   async forgetMachine(mac: string): Promise<{ status: string }> {
     return this.request("DELETE", `/api/machines/${encodeURIComponent(mac)}`);
   }

bastion/src/cli/src/commands/app.ts

@@ -70,7 +70,7 @@ export function registerAppCommand(program: Command): void {
     .command("install <target>")
     .description("Install k3s on a target machine (hostname, IP, or MAC)")
     .option("--role <role>", "k3s role: infra (server) or worker (agent)", "infra")
-    .option("--user <user>", "SSH user", "michal")
+    .option("--user <user>", "SSH user", "root")
     .option("--k3s-server <url>", "k3s server URL (required for worker role)")
     .option("--k3s-token <token>", "k3s join token (required for worker role)")
     .action(async (target: string, opts: {
@@ -164,7 +164,7 @@ export function registerAppCommand(program: Command): void {
   k3sCmd
     .command("health [target]")
     .description("Check k3s health (all hosts if no target given)")
-    .option("--user <user>", "SSH user", "michal")
+    .option("--user <user>", "SSH user", "root")
     .action(async (target: string | undefined, opts: { user: string }) => {
       const sshKey = findSshKey();
 
@@ -304,7 +304,7 @@ export function registerAppCommand(program: Command): void {
   k3sCmd
     .command("list")
     .description("List installed machines and their k3s status")
-    .option("--user <user>", "SSH user", "michal")
+    .option("--user <user>", "SSH user", "root")
     .action(async (opts: { user: string }) => {
       let state: BastionState;
       try {

bastion/src/cli/src/commands/asahi.ts

@@ -0,0 +1,69 @@
+// CLI command: provision asahi
+// Prints the curl command to run on the Mac Studio (macOS) to install
+// Fedora Asahi Remix with lab LVM layout.
+
+import type { Command } from "commander";
+import { getLabdClient } from "../api/config.js";
+
+export function registerAsahiCommand(parent: Command): void {
+  parent
+    .command("asahi")
+    .description("Show instructions to provision an Apple Silicon Mac with Asahi Linux")
+    .action(async () => {
+      // Try to get bastion info to determine the correct URL
+      let bastionUrl = "";
+      try {
+        const bastions = await getLabdClient().getBastions();
+        const online = bastions.find(b => b.status === "online");
+        if (online) {
+          bastionUrl = `http://${online.serverIp}:8080`;
+        }
+      } catch { /* labd not reachable */ }
+
+      if (!bastionUrl) {
+        // Fall back to config
+        const { loadConfig } = await import("../config/index.js");
+        const config = loadConfig();
+        bastionUrl = config.labdUrl ?? "http://<bastion-ip>:8080";
+        // Convert labd URL to bastion URL (labd is on different port/host)
+        bastionUrl = bastionUrl.replace(/:\d+$/, ":8080");
+      }
+
+      const BOLD = "\x1b[1m";
+      const CYAN = "\x1b[36m";
+      const DIM = "\x1b[2m";
+      const RESET = "\x1b[0m";
+
+      console.log("");
+      console.log(`${BOLD}  Asahi Linux Provisioning${RESET}`);
+      console.log(`${DIM}  For Apple Silicon Macs (Mac Studio, MacBook, etc.)${RESET}`);
+      console.log("");
+      console.log(`  Run this command ${BOLD}on the Mac${RESET} (from macOS Terminal):`);
+      console.log("");
+      console.log(`    ${CYAN}${BOLD}curl ${bastionUrl}/asahi | sh${RESET}`);
+      console.log("");
+      console.log(`  The installer will ask a few interactive questions:`);
+      console.log(`    ${BOLD}1.${RESET} Action: press ${BOLD}r${RESET} to resize macOS`);
+      console.log(`    ${BOLD}2.${RESET} How much space for Linux: choose maximum`);
+      console.log(`    ${BOLD}3.${RESET} Confirm the resize operation`);
+      console.log(`    ${BOLD}4.${RESET} macOS password for firmware authentication`);
+      console.log("");
+      console.log(`  After that, everything is automatic:`);
+      console.log(`    - Asahi boot infrastructure (m1n1 + U-Boot)`);
+      console.log(`    - Fedora Asahi Remix root partition`);
+      console.log(`    - LVM data partition (remaining space)`);
+      console.log("");
+      console.log(`  On first boot, LVM volumes are created automatically:`);
+      console.log(`    ${DIM}labvg/swap (27GB), labvg/var (100GB), labvg/varlog (10GB),`);
+      console.log(`    labvg/home (10GB), labvg/srv (20GB), labvg/rancher (20GB),`);
+      console.log(`    labvg/longhorn (remaining space)${RESET}`);
+      console.log("");
+      console.log(`  After first boot, SSH in and run the firstboot script:`);
+      console.log(`    ${BOLD}ssh root@<ip> 'curl -sf ${bastionUrl}/asahi/firstboot.sh | bash'${RESET}`);
+      console.log("");
+      console.log(`  This sets up LVM, detects hostname/MAC, and self-registers.`);
+      console.log(`  Then install k3s:`);
+      console.log(`    ${BOLD}labctl app k3s install <hostname> --role infra${RESET}`);
+      console.log("");
+    });
+}

bastion/src/cli/src/commands/labcontroller.ts

@@ -38,7 +38,7 @@ export function registerLabcontrollerCommands(appCmd: Command): void {
   lcCmd
     .command("deploy <target>")
     .description("Deploy labcontroller stack to a k3s node")
-    .option("--user <user>", "SSH user", "michal")
+    .option("--user <user>", "SSH user", "root")
     .option("--crdb-replicas <n>", "CockroachDB replicas", "1")
     .action(async (target: string, opts: {
       user: string;
@@ -193,7 +193,7 @@ export function registerLabcontrollerCommands(appCmd: Command): void {
   lcCmd
     .command("status [target]")
     .description("Check labcontroller deployment status (all hosts if no target)")
-    .option("--user <user>", "SSH user", "michal")
+    .option("--user <user>", "SSH user", "root")
     .action(async (target: string | undefined, opts: { user: string }) => {
       const sshKey = findSshKey();
       const sshOpts = sshKey ? { keyPath: sshKey } : {};

bastion/src/cli/src/commands/list.ts

@@ -69,10 +69,10 @@ export function registerListCommand(parent: Command): void {
         const hostname = inst?.hostname ?? queued?.hostname ?? "-";
         const role = inst?.role ?? queued?.role ?? "-";
         const ip = inst?.ip ?? "-";
-        const cpu = hw?.cpu_model ?? "-";
+        const cpu = hw?.cpu_model ?? inst?.cpu_model ?? "-";
-        const cores = hw?.cpu_cores != null ? String(hw.cpu_cores) : "-";
+        const cores = (hw?.cpu_cores ?? inst?.cpu_cores) != null ? String(hw?.cpu_cores ?? inst?.cpu_cores) : "-";
-        const ram = hw?.memory_gb != null ? `${hw.memory_gb}GB` : "-";
+        const ram = (hw?.memory_gb ?? inst?.memory_gb) != null ? `${hw?.memory_gb ?? inst?.memory_gb}GB` : "-";
-        const product = hw?.product ?? "-";
+        const product = hw?.product ?? inst?.product ?? "-";
 
         const color = statusColor(status);
 

bastion/src/cli/src/commands/recheck.ts

@@ -0,0 +1,94 @@
+// CLI command: provision recheck
+// SSH into all installed machines, collect hardware info, update bastion state.
+
+import type { Command } from "commander";
+import { sshExec } from "@lab/modules";
+import { getLabdClient } from "../api/config.js";
+
+const BOLD = "\x1b[1m";
+const GREEN = "\x1b[0;32m";
+const RED = "\x1b[0;31m";
+const DIM = "\x1b[2m";
+const RESET = "\x1b[0m";
+
+const SSH_OPTS = { timeoutMs: 30_000 };
+
+// Shell script that collects hardware info as JSON.
+// Kept simple — no Python, pure shell + awk.
+const HW_COLLECT_SCRIPT = [
+  'P=$(cat /sys/class/dmi/id/product_name 2>/dev/null || echo unknown)',
+  'B=$(cat /sys/class/dmi/id/board_name 2>/dev/null || echo unknown)',
+  'S=$(cat /sys/class/dmi/id/product_serial 2>/dev/null || echo unknown)',
+  'M=$(cat /sys/class/dmi/id/sys_vendor 2>/dev/null || echo unknown)',
+  'C=$(grep -m1 "model name" /proc/cpuinfo 2>/dev/null | cut -d: -f2 | sed "s/^ //" || grep -m1 Model /proc/cpuinfo 2>/dev/null | cut -d: -f2 | sed "s/^ //" || echo unknown)',
+  'N=$(grep -c "^processor" /proc/cpuinfo 2>/dev/null || echo 0)',
+  'R=$(awk "/MemTotal/ {printf \\"%d\\", \\$2/1024/1024}" /proc/meminfo 2>/dev/null || echo 0)',
+  'A=$(uname -m)',
+  'printf \'{"product":"%s","board":"%s","serial":"%s","manufacturer":"%s","cpu_model":"%s","cpu_cores":%s,"memory_gb":%s,"arch":"%s"}\\n\' "$P" "$B" "$S" "$M" "$C" "$N" "$R" "$A"',
+].join("; ");
+
+export function registerRecheckCommand(parent: Command): void {
+  parent
+    .command("recheck")
+    .description("Refresh hardware info for all installed machines via SSH")
+    .option("--user <user>", "SSH user", "root")
+    .option("--target <hostname>", "Only recheck a specific machine (by hostname or MAC)")
+    .action(async (opts: { user: string; target?: string }) => {
+      const client = getLabdClient();
+      let state;
+      try {
+        state = await client.getMachines();
+      } catch (err) {
+        console.error(`Cannot reach labd: ${err instanceof Error ? err.message : String(err)}`);
+        process.exit(1);
+      }
+
+      // Build list of machines to check
+      const targets: Array<{ mac: string; hostname: string; ip: string }> = [];
+      for (const [mac, info] of Object.entries(state.installed)) {
+        if (!info.ip) continue;
+        if (opts.target && info.hostname !== opts.target && mac !== opts.target) continue;
+        targets.push({ mac, hostname: info.hostname, ip: info.ip });
+      }
+
+      if (targets.length === 0) {
+        console.log("No installed machines with IPs to check.");
+        return;
+      }
+
+      console.log(`\n${BOLD}Rechecking ${targets.length} machine(s)...${RESET}\n`);
+
+      let updated = 0;
+      let failed = 0;
+
+      for (const { mac, hostname, ip } of targets) {
+        process.stdout.write(`  ${hostname.padEnd(24)} ${DIM}(${ip})${RESET}  `);
+
+        try {
+          const t0 = Date.now();
+          const result = await sshExec(ip, opts.user, HW_COLLECT_SCRIPT, SSH_OPTS);
+          const elapsed = Date.now() - t0;
+          if (result.exitCode !== 0) {
+            console.log(`${RED}SSH failed (exit ${result.exitCode}, ${elapsed}ms)${RESET}`);
+            if (result.stderr) console.log(`    ${DIM}${result.stderr.substring(0, 200)}${RESET}`);
+            console.log(`${RED}SSH failed (exit ${result.exitCode})${RESET}`);
+            failed++;
+            continue;
+          }
+
+          const hwData = JSON.parse(result.stdout.trim());
+          await client.discoverMachine({ mac, ...hwData });
+          const cpu = hwData.cpu_model || "?";
+          const cores = hwData.cpu_cores || "?";
+          const mem = hwData.memory_gb || "?";
+          console.log(`${GREEN}OK${RESET} ${DIM}${cpu}, ${cores} cores, ${mem}GB${RESET}`);
+          updated++;
+        } catch (err) {
+          console.log(`${RED}FAIL${RESET} ${DIM}${err instanceof Error ? err.message : String(err)}${RESET}`);
+          failed++;
+        }
+      }
+
+      console.log(`\n${BOLD}Done:${RESET} ${updated} updated, ${failed} failed\n`);
+    });
+}

bastion/src/cli/src/index.ts

@@ -17,8 +17,10 @@ import { registerReprovisionCommand } from "./commands/reprovision.js";
 import { registerDebugCommand } from "./commands/debug.js";
 import { registerForgetCommand } from "./commands/forget.js";
 import { registerRegisterCommand } from "./commands/register.js";
+import { registerAsahiCommand } from "./commands/asahi.js";
 import { registerLogsCommand } from "./commands/logs.js";
 import { registerMakeIsoCommand } from "./commands/makeiso.js";
+import { registerRecheckCommand } from "./commands/recheck.js";
 import { registerConfigCommand } from "./commands/config.js";
 import { registerLoginCommand } from "./commands/login.js";
 import { registerDoctorCommand } from "./commands/doctor.js";
@@ -100,8 +102,10 @@ export function createProgram(): Command {
   registerDebugCommand(provisionCmd);
   registerForgetCommand(provisionCmd);
   registerRegisterCommand(provisionCmd);
+  registerAsahiCommand(provisionCmd);
   registerLogsCommand(provisionCmd);
   registerMakeIsoCommand(provisionCmd);
+  registerRecheckCommand(provisionCmd);
 
   // config list/get/set/path
   registerConfigCommand(program);

bastion/src/cli/tests/smoke-bastion.test.ts

@@ -137,7 +137,7 @@ describe("bastion smoke tests", () => {
 
     // Wait for the server to start (look for the banner)
     const startedAt = Date.now();
-    const maxWait = 10_000;
+    const maxWait = 15_000;
     while (Date.now() - startedAt < maxWait) {
       if (stdout.includes("Waiting for PXE boot requests")) break;
       await sleep(200);

bastion/src/labd/src/routes/bastions.ts

@@ -260,6 +260,37 @@ export function registerBastionRoutes(app: FastifyInstance, db: DbClient): void
     }
   });
 
+  // Update hardware info (discovery data) for a machine
+  app.post<{
+    Body: {
+      mac?: string; product?: string; board?: string; serial?: string;
+      manufacturer?: string; cpu_model?: string; cpu_cores?: number;
+      memory_gb?: number; arch?: string;
+      disks?: Array<{ name: string; size_gb: number; model: string }>;
+      nics?: Array<{ name: string; mac: string; state: string }>;
+    };
+  }>("/api/machines/discover", async (request, reply) => {
+    const data = request.body ?? {};
+    const mac = (data.mac ?? "").toLowerCase().replace(/-/g, ":");
+    if (!mac) {
+      return reply.code(400).send({ error: "mac is required" });
+    }
+
+    const bastion = bastionRegistry.findBastionByMac(mac);
+    const target = bastion ?? (bastionRegistry.getAll().length === 1 ? bastionRegistry.getAll()[0] : null);
+
+    if (!target) {
+      return reply.code(503).send({ error: "No bastion found for this MAC" });
+    }
+
+    try {
+      const result = await sendCommand(target.bastionId, { type: "command-discover", ...data, mac });
+      return reply.code(result.status === "ok" ? 200 : 500).send(result);
+    } catch (err) {
+      return reply.code(500).send({ error: err instanceof Error ? err.message : String(err) });
+    }
+  });
+
   // Update role
   app.post<{
     Body: { mac?: string; role?: string };

bastion/src/modules/modules/k3s/src/groups/hardening.ts

@@ -1,18 +1,22 @@
-// Hardening: Pod Security Standards, certificate check, log rotation.
+// Hardening: Pod Security Standards, certificate check, journald cap, storage.
 
 import type { OperationContext, OperationResult, OperationGroup } from "../types.js";
 import { runSequential } from "../utils.js";
 import { applyPodSecurityStandards } from "../operations/pod-security.js";
 import { checkCertExpiry } from "../operations/cert-check.js";
 import { configureLogRotation } from "../operations/log-rotation.js";
+import { configureJournaldLimits } from "../operations/journald-limits.js";
+import { configureLonghornDisk } from "../operations/longhorn-disk.js";
 
 export const hardeningGroup: OperationGroup = {
   name: "hardening",
-  description: "Pod security, certificate check, log rotation",
+  description: "Pod security, certificate check, journald cap, storage",
   operations: [
     { name: "Apply Pod Security Standards", fn: applyPodSecurityStandards },
     { name: "Check certificate expiry", fn: checkCertExpiry },
-    { name: "Configure log rotation", fn: configureLogRotation },
+    { name: "Decommission file-based audit logs", fn: configureLogRotation },
+    { name: "Configure journald disk cap", fn: configureJournaldLimits },
+    { name: "Configure Longhorn disk", fn: configureLonghornDisk },
   ],
 };
 

bastion/src/modules/modules/k3s/src/groups/host-prep.ts

@@ -7,16 +7,18 @@ import { applyCisHardening } from "../operations/sysctl.js";
 import { disableSwap } from "../operations/swap.js";
 import { disableFirewall } from "../operations/firewall.js";
 import { setSelinuxPermissive } from "../operations/selinux.js";
+import { enableIscsi } from "../operations/iscsi.js";
 
 export const hostPrepGroup: OperationGroup = {
   name: "host-prep",
-  description: "Prepare host for k3s: kernel modules, sysctl, swap, firewall, SELinux",
+  description: "Prepare host for k3s: kernel modules, sysctl, swap, firewall, SELinux, iSCSI",
   operations: [
     { name: "Load kernel modules", fn: loadKernelModules },
     { name: "Apply CIS sysctl", fn: applyCisHardening },
     { name: "Disable swap", fn: disableSwap },
     { name: "Disable firewall", fn: disableFirewall },
     { name: "Set SELinux permissive", fn: setSelinuxPermissive },
+    { name: "Enable iSCSI", fn: enableIscsi },
   ],
 };
 

bastion/src/modules/modules/k3s/src/install.ts

@@ -76,7 +76,6 @@ sed -i 's/^SELINUX=enforcing/SELINUX=permissive/' /etc/selinux/config 2>/dev/nul
 # ── 5b. Create k3s config directory ──
 echo "[5/10] Writing k3s server configuration..."
 mkdir -p /etc/rancher/k3s
-mkdir -p /var/log/kubernetes
 
 cat > /etc/rancher/k3s/config.yaml << 'K3S_CONFIG'
 # k3s server configuration — CIS hardened
@@ -91,13 +90,10 @@ disable:
   - servicelb
   - traefik
 
-# API server hardening
+# API server hardening (audit-log-path=- routes audit to journald via stdout)
 kube-apiserver-arg:
   - "anonymous-auth=false"
-  - "audit-log-path=/var/log/kubernetes/audit.log"
+  - "audit-log-path=-"
-  - "audit-log-maxage=30"
-  - "audit-log-maxbackup=10"
-  - "audit-log-maxsize=100"
   - "audit-policy-file=/etc/rancher/k3s/audit-policy.yaml"
   - "enable-admission-plugins=NodeRestriction,PodSecurity"
   - "request-timeout=300s"

bastion/src/modules/modules/k3s/src/k3s-module.ts

@@ -78,9 +78,10 @@ export class K3sModule implements Module {
       return toModuleResult("install", [...prepResults, ...k3sResults], start);
     }
 
-    // Phase 3: Networking (server only — agents don't install Cilium)
+    // Phase 3: Networking (initial server only — joining servers get Cilium via daemonset)
     let netResults: OperationResult[] = [];
-    if (isServer) {
+    const isJoiningServer = isServer && !!opCtx.config.k3sServerUrl;
+    if (isServer && !isJoiningServer) {
       netResults = await runNetworking(opCtx);
     }
 

bastion/src/modules/modules/k3s/src/operations/etcd-recover.ts

@@ -0,0 +1,194 @@
+// Recover a broken etcd member by removing it from the cluster, wiping its
+// local state, and restarting k3s so it rejoins as a fresh member.
+//
+// Use case: a node panics on startup with
+//   "tocommit(N+1) is out of range [lastIndex(N)]. Was the raft log corrupted,
+//    truncated, or lost?"
+// This means the local raft WAL is missing the last entry the leader thinks
+// the follower acknowledged (lost write, unclean shutdown, etc). The fix is
+// always the same and well-documented; this codifies it so we don't fumble
+// the procedure under pressure.
+//
+// Preconditions:
+//   - At least one healthy peer is reachable so the cluster has quorum after
+//     we remove the broken member. (For a 3-node cluster: 2 healthy. For a
+//     5-node: 3 healthy.) If quorum would be lost, this function refuses.
+//   - SSH access to both the broken node and a healthy peer.
+//   - etcdctl available on the healthy peer (k3s does not bundle it; the
+//     procedure installs it on demand on Fedora).
+
+import type { SshClient } from "../types.js";
+
+const ETCD_TLS = {
+  ca: "/var/lib/rancher/k3s/server/tls/etcd/server-ca.crt",
+  cert: "/var/lib/rancher/k3s/server/tls/etcd/server-client.crt",
+  key: "/var/lib/rancher/k3s/server/tls/etcd/server-client.key",
+} as const;
+
+const SSH_TIMEOUT = 60_000;
+
+export interface RecoverEtcdMemberOptions {
+  /** SSH client for the broken node (the one panicking). */
+  broken: SshClient;
+  /** SSH client for any healthy server peer in the same cluster. */
+  peer: SshClient;
+  /** Hostname (k8s node name) of the broken node. Used to find its etcd member id. */
+  brokenHostname: string;
+  /** Logger for progress output. */
+  log?: (msg: string) => void;
+}
+
+export interface RecoverEtcdMemberResult {
+  success: boolean;
+  changed: boolean;
+  message: string;
+  /** New etcd member id assigned after rejoin (when known). */
+  newMemberId?: string;
+  /** Old etcd member id that was removed. */
+  removedMemberId?: string;
+  error?: string;
+}
+
+function etcdctl(subcmd: string): string {
+  return [
+    "ETCDCTL_API=3 etcdctl",
+    `--cacert=${ETCD_TLS.ca}`,
+    `--cert=${ETCD_TLS.cert}`,
+    `--key=${ETCD_TLS.key}`,
+    "--endpoints=https://127.0.0.1:2379",
+    "--command-timeout=10s",
+    subcmd,
+  ].join(" ");
+}
+
+async function ensureEtcdctl(peer: SshClient): Promise<void> {
+  const probe = await peer.exec("command -v etcdctl 2>/dev/null", { timeoutMs: 5_000 });
+  if (probe.exitCode === 0 && probe.stdout.trim()) return;
+  // Best-effort install on Fedora. If the host isn't dnf-based, surface the
+  // error to the caller via the next etcdctl invocation.
+  await peer.exec("dnf install -y etcd 2>&1", { timeoutMs: 120_000 });
+}
+
+async function getMemberList(peer: SshClient): Promise<Array<{ id: string; name: string }>> {
+  const result = await peer.exec(etcdctl("member list"), { timeoutMs: SSH_TIMEOUT });
+  if (result.exitCode !== 0) {
+    throw new Error(`etcdctl member list failed: ${result.stderr || result.stdout}`);
+  }
+  // Format: <hex-id>, started, <name>, <peer-urls>, <client-urls>, <isLearner>
+  return result.stdout
+    .split("\n")
+    .map((line) => line.trim())
+    .filter(Boolean)
+    .map((line) => {
+      const [id, , name] = line.split(",").map((p) => p.trim());
+      return { id: id ?? "", name: name ?? "" };
+    })
+    .filter((m) => m.id);
+}
+
+export async function recoverEtcdMember(
+  opts: RecoverEtcdMemberOptions,
+): Promise<RecoverEtcdMemberResult> {
+  const log = opts.log ?? (() => {});
+
+  try {
+    log(`Looking up etcd member id for ${opts.brokenHostname} via peer...`);
+    await ensureEtcdctl(opts.peer);
+
+    const members = await getMemberList(opts.peer);
+    if (members.length < 3) {
+      return {
+        success: false,
+        changed: false,
+        message: "Refusing to remove a member from a cluster with <3 members (quorum would be lost)",
+        error: `member count = ${members.length}`,
+      };
+    }
+
+    // Member names are <hostname>-<random-suffix>; match by hostname prefix.
+    const broken = members.find((m) => m.name.startsWith(opts.brokenHostname));
+    if (!broken) {
+      return {
+        success: false,
+        changed: false,
+        message: `No etcd member found matching hostname ${opts.brokenHostname}`,
+        error: `members: ${members.map((m) => m.name).join(", ")}`,
+      };
+    }
+    log(`Broken member: ${broken.id} (${broken.name})`);
+
+    log("Step 1/4: stopping k3s on broken node");
+    await opts.broken.exec("systemctl stop k3s 2>&1", { timeoutMs: SSH_TIMEOUT });
+
+    log("Step 2/4: removing broken etcd member from cluster");
+    const remove = await opts.peer.exec(
+      etcdctl(`member remove ${broken.id}`),
+      { timeoutMs: SSH_TIMEOUT },
+    );
+    if (remove.exitCode !== 0) {
+      return {
+        success: false,
+        changed: false,
+        message: "etcdctl member remove failed",
+        error: remove.stderr || remove.stdout,
+        removedMemberId: broken.id,
+      };
+    }
+
+    log("Step 3/4: archiving corrupt etcd state and stale TLS/cred dirs on broken node");
+    const ts = Math.floor(Date.now() / 1000);
+    await opts.broken.exec(
+      [
+        `mv /var/lib/rancher/k3s/server/db /var/lib/rancher/k3s/server/db.corrupt-${ts} 2>/dev/null || true`,
+        "rm -rf /var/lib/rancher/k3s/server/tls /var/lib/rancher/k3s/server/cred",
+      ].join(" && "),
+      { timeoutMs: SSH_TIMEOUT },
+    );
+
+    log("Step 4/4: starting k3s on broken node — it will rejoin");
+    await opts.broken.exec("systemctl start k3s 2>&1", { timeoutMs: SSH_TIMEOUT });
+
+    // Poll for rejoin. The new member-id is what the cluster assigns on join.
+    let newMemberId: string | undefined;
+    for (let i = 0; i < 60; i++) {
+      await new Promise((r) => setTimeout(r, 5_000));
+      try {
+        const after = await getMemberList(opts.peer);
+        const rejoined = after.find(
+          (m) => m.name.startsWith(opts.brokenHostname) && m.id !== broken.id,
+        );
+        if (rejoined) {
+          newMemberId = rejoined.id;
+          break;
+        }
+      } catch {
+        // peer may briefly be unreachable mid-rejoin — keep polling
+      }
+    }
+
+    if (!newMemberId) {
+      return {
+        success: false,
+        changed: true,
+        message: "k3s started but new member did not appear in cluster within 5 minutes",
+        removedMemberId: broken.id,
+      };
+    }
+
+    log(`Rejoined as ${newMemberId}`);
+    return {
+      success: true,
+      changed: true,
+      message: `Recovered: removed ${broken.id}, rejoined as ${newMemberId}`,
+      removedMemberId: broken.id,
+      newMemberId,
+    };
+  } catch (err) {
+    return {
+      success: false,
+      changed: false,
+      message: "Recovery failed",
+      error: err instanceof Error ? err.message : String(err),
+    };
+  }
+}

bastion/src/modules/modules/k3s/src/operations/index.ts

@@ -1,6 +1,7 @@
 export { loadKernelModules } from "./kernel-modules.js";
 export { applyCisHardening } from "./sysctl.js";
 export { disableSwap } from "./swap.js";
+export { enableIscsi } from "./iscsi.js";
 export { disableFirewall } from "./firewall.js";
 export { setSelinuxPermissive } from "./selinux.js";
 export { writeK3sConfig } from "./k3s-config.js";
@@ -10,6 +11,13 @@ export { installK3sBinary } from "./k3s-install.js";
 export { installCilium } from "./cilium.js";
 export { fixCoreDnsUpstream } from "./dns-fix.js";
 export { configureLogRotation } from "./log-rotation.js";
+export { configureJournaldLimits } from "./journald-limits.js";
 export { applyDefaultNetworkPolicies } from "./network-policy.js";
 export { applyPodSecurityStandards } from "./pod-security.js";
 export { checkCertExpiry } from "./cert-check.js";
+export { configureLonghornDisk } from "./longhorn-disk.js";
+export { recoverEtcdMember } from "./etcd-recover.js";
+export type {
+  RecoverEtcdMemberOptions,
+  RecoverEtcdMemberResult,
+} from "./etcd-recover.js";

bastion/src/modules/modules/k3s/src/operations/iscsi.ts

@@ -0,0 +1,31 @@
+// Install and enable iSCSI initiator (required by Longhorn storage).
+// Fedora: iscsi-initiator-utils, Ubuntu: open-iscsi
+
+import type { Operation, OperationResult } from "../types.js";
+import { sshOpts } from "../utils.js";
+
+export const enableIscsi: Operation = async (ctx): Promise<OperationResult> => {
+  // Check if iscsid is already running
+  const check = await ctx.ssh.exec("systemctl is-active iscsid 2>/dev/null", sshOpts(ctx));
+  if (check.stdout.trim() === "active") {
+    return { success: true, changed: false, message: "iSCSI already active" };
+  }
+
+  // Install the package (detect distro)
+  const osRelease = await ctx.ssh.exec("cat /etc/os-release", sshOpts(ctx));
+  const osLower = osRelease.stdout.toLowerCase();
+  const isFedora = osLower.includes("fedora") || osLower.includes("rhel") || osLower.includes("centos");
+
+  const pkg = isFedora ? "iscsi-initiator-utils" : "open-iscsi";
+  const installCmd = isFedora ? `sudo dnf install -y ${pkg}` : `sudo apt-get install -y ${pkg}`;
+
+  const install = await ctx.ssh.exec(installCmd, { timeoutMs: 120_000 });
+  if (install.exitCode !== 0) {
+    return { success: false, changed: false, message: `Failed to install ${pkg}`, error: install.stderr.trim() };
+  }
+
+  // Enable and start
+  await ctx.ssh.exec("sudo systemctl enable --now iscsid", sshOpts(ctx));
+
+  return { success: true, changed: true, message: `Installed ${pkg} and enabled iscsid` };
+};

bastion/src/modules/modules/k3s/src/operations/journald-limits.ts

@@ -0,0 +1,33 @@
+// Cap journald disk usage so audit logs (which now flow through journald via
+// kube-apiserver's stdout) cannot fill /var/log. Default journald uses up to
+// 10% of the filesystem, capped at 4 GB. In a /var/log of ~10 GB shared with
+// other services, that's still room for audit volume to evict useful logs.
+// 2 GB / 200 MB-per-file is a comfortable middle.
+
+import type { Operation, OperationResult } from "../types.js";
+import { sshOpts, writeRemoteFile } from "../utils.js";
+
+const DROPIN_CONTENT = `[Journal]
+SystemMaxUse=2G
+SystemKeepFree=1G
+SystemMaxFileSize=200M
+`;
+
+const DROPIN_PATH = "/etc/systemd/journald.conf.d/10-k3s-audit-cap.conf";
+
+export const configureJournaldLimits: Operation = async (ctx): Promise<OperationResult> => {
+  const changed = await writeRemoteFile(ctx, DROPIN_PATH, DROPIN_CONTENT);
+  if (changed) {
+    // Reload journald so the new limit applies without a reboot.
+    await ctx.ssh.exec(
+      "systemctl kill --signal=SIGUSR2 systemd-journald 2>/dev/null; " +
+      "systemctl restart systemd-journald 2>&1 || true",
+      sshOpts(ctx),
+    );
+  }
+  return {
+    success: true,
+    changed,
+    message: changed ? "journald limits configured (2 GB cap)" : "journald limits already configured",
+  };
+};

bastion/src/modules/modules/k3s/src/operations/k3s-config.ts

@@ -9,7 +9,18 @@ function isServerRole(role: string): boolean {
 
 function generateServerConfig(config: K3sConfig): string {
   const tlsSans = [config.hostname, config.ip, ...(config.tlsSans ?? [])];
-  return `# k3s server configuration — CIS hardened
+  const isJoining = !!config.k3sServerUrl;
+  const clusterLines = isJoining
+    ? `server: "${config.k3sServerUrl}"\ntoken: "${config.k3sToken}"`
+    : "cluster-init: true";
+  // audit-log-path=- routes audit events to k3s.service's stdout, which systemd
+  // forwards to journald. journald enforces its own size caps (see
+  // configureJournaldLimits) so audit volume cannot fill the disk. File-based
+  // audit logs led to /var/log/kubernetes growing to 7+ GB because apiserver's
+  // own rotation produced files that any logrotate glob would double-rotate
+  // and never expire.
+  return `# k3s server configuration — CIS hardened, etcd HA
+${clusterLines}
 protect-kernel-defaults: true
 secrets-encryption: true
 write-kubeconfig-mode: "0640"
@@ -20,12 +31,12 @@ disable:
   - servicelb
   - traefik
 
+node-label:
+  - "node.longhorn.io/create-default-disk=config"
+
 kube-apiserver-arg:
   - "anonymous-auth=false"
-  - "audit-log-path=/var/log/kubernetes/audit.log"
+  - "audit-log-path=-"
-  - "audit-log-maxage=30"
-  - "audit-log-maxbackup=10"
-  - "audit-log-maxsize=100"
   - "audit-policy-file=/etc/rancher/k3s/audit-policy.yaml"
   - "enable-admission-plugins=NodeRestriction,PodSecurity"
   - "request-timeout=300s"
@@ -44,6 +55,7 @@ function generateAgentConfig(): string {
   return `protect-kernel-defaults: true
 node-label:
   - "node-role.kubernetes.io/worker=true"
+  - "node.longhorn.io/create-default-disk=config"
 kubelet-arg:
   - "protect-kernel-defaults=true"
   - "streaming-connection-idle-timeout=5m"
@@ -52,7 +64,7 @@ kubelet-arg:
 }
 
 export const writeK3sConfig: Operation = async (ctx): Promise<OperationResult> => {
-  await ctx.ssh.exec("mkdir -p /etc/rancher/k3s /var/log/kubernetes", sshOpts(ctx));
+  await ctx.ssh.exec("mkdir -p /etc/rancher/k3s", sshOpts(ctx));
 
   const content = isServerRole(ctx.config.role)
     ? generateServerConfig(ctx.config)

bastion/src/modules/modules/k3s/src/operations/k3s-install.ts

@@ -15,8 +15,21 @@ export const installK3sBinary: Operation = async (ctx): Promise<OperationResult>
   const alreadyInstalled = version.exitCode === 0;
 
   if (isServer) {
+    // Clean stale server state when joining an existing cluster
+    // (TLS certs from a previous run cause "newer than datastore" fatal error)
+    if (ctx.config.k3sServerUrl && ctx.config.k3sToken) {
+      await ctx.ssh.exec(
+        "rm -rf /var/lib/rancher/k3s/server/tls /var/lib/rancher/k3s/server/cred /var/lib/rancher/k3s/server/db",
+        sshOpts(ctx),
+      );
+    }
+
+    // If joining an existing cluster, pass K3S_URL and K3S_TOKEN
+    const joinEnv = ctx.config.k3sServerUrl && ctx.config.k3sToken
+      ? `K3S_URL="${ctx.config.k3sServerUrl}" K3S_TOKEN="${ctx.config.k3sToken}"`
+      : "";
     const result = await ctx.ssh.exec(
-      'curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="server" INSTALL_K3S_SKIP_SELINUX_RPM=true sh -',
+      `curl -sfL https://get.k3s.io | INSTALL_K3S_EXEC="server" INSTALL_K3S_SKIP_SELINUX_RPM=true ${joinEnv} sh -`,
       { timeoutMs: 300_000 },
     );
     if (result.exitCode !== 0) {

bastion/src/modules/modules/k3s/src/operations/log-rotation.ts

@@ -1,25 +1,44 @@
-// Configure log rotation for k3s.
+// Decommission file-based k8s audit logging in favor of journald.
+//
+// Earlier versions wrote audit events to /var/log/kubernetes/audit.log and
+// rotated them with a logrotate rule. Two failure modes followed: kube-apiserver
+// rotated internally (audit-{ts}.log), the *.log glob in logrotate
+// double-rotated those (-{date}), and the resulting filename matched no
+// retention policy, so the directory grew unbounded (we observed 7+ GB).
+//
+// k3s now sets audit-log-path=- so audit goes to stdout → journald, which
+// enforces SystemMaxUse caps. This operation removes the obsolete logrotate
+// rule and reaps any audit files left behind by the old setup. Idempotent: on
+// fresh installs everything is already absent and the operation is a no-op.
 
 import type { Operation, OperationResult } from "../types.js";
-import { writeRemoteFile } from "../utils.js";
+import { sshOpts } from "../utils.js";
 
-const LOGROTATE_CONFIG = `/var/log/kubernetes/*.log {
+const REMOVE_LOGROTATE = "rm -f /etc/logrotate.d/k3s";
-  daily
+
-  rotate 14
+// Bounded by a max-depth and explicit name pattern so we never reach outside
-  compress
+// the deprecated audit-log directory.
-  delaycompress
+const REAP_OLD_AUDIT_FILES =
-  missingok
+  "find /var/log/kubernetes -maxdepth 1 -type f " +
-  notifempty
+  "\\( -name 'audit*.log*' -o -name 'audit-*.log' \\) " +
-  copytruncate
+  "-delete 2>/dev/null; " +
-  maxsize 100M
+  "rmdir /var/log/kubernetes 2>/dev/null; true";
-}`;
 
 export const configureLogRotation: Operation = async (ctx): Promise<OperationResult> => {
-  const changed = await writeRemoteFile(ctx, "/etc/logrotate.d/k3s", LOGROTATE_CONFIG);
+  const before = await ctx.ssh.exec(
+    "test -e /etc/logrotate.d/k3s -o -d /var/log/kubernetes && echo present || echo absent",
+    sshOpts(ctx),
+  );
+  const wasPresent = before.stdout.trim() === "present";
+
+  await ctx.ssh.exec(REMOVE_LOGROTATE, sshOpts(ctx));
+  await ctx.ssh.exec(REAP_OLD_AUDIT_FILES, sshOpts(ctx));
 
   return {
     success: true,
-    changed,
+    changed: wasPresent,
-    message: changed ? "Log rotation configured" : "Log rotation already configured",
+    message: wasPresent
+      ? "Removed legacy file-based audit logging (now via journald)"
+      : "No legacy audit log artifacts present",
   };
 };

bastion/src/modules/modules/k3s/src/operations/longhorn-disk.ts

@@ -0,0 +1,50 @@
+// Annotate nodes with Longhorn default disk config when /var/lib/longhorn exists.
+// The label is set in k3s config (node-label), but the annotation must be applied via kubectl.
+
+import type { Operation, OperationResult } from "../types.js";
+import { sshOpts } from "../utils.js";
+import { sshExec as remoteSshExec } from "../../../../src/ssh.js";
+
+export const configureLonghornDisk: Operation = async (ctx): Promise<OperationResult> => {
+  // Check if /var/lib/longhorn exists on this node
+  const check = await ctx.ssh.exec("test -d /var/lib/longhorn && echo yes || echo no", sshOpts(ctx));
+  if (check.stdout.trim() !== "yes") {
+    return { success: true, changed: false, message: "No /var/lib/longhorn directory — skipping Longhorn disk config" };
+  }
+
+  // Find the node name (hostname as registered in k3s)
+  const nodeNameResult = await ctx.ssh.exec("hostname -f 2>/dev/null || hostname", sshOpts(ctx));
+  const nodeName = nodeNameResult.stdout.trim();
+
+  const annotation = JSON.stringify([{ path: "/var/lib/longhorn", allowScheduling: true }]);
+
+  // Try kubectl locally first (works on server nodes)
+  const result = await ctx.ssh.exec(
+    `k3s kubectl annotate node "${nodeName}" "node.longhorn.io/default-disks-config=${annotation}" --overwrite 2>&1 || true`,
+    sshOpts(ctx),
+  );
+
+  if (result.stdout.includes("annotated") || result.stdout.includes("unchanged")) {
+    return { success: true, changed: true, message: `Longhorn disk annotation applied to ${nodeName}` };
+  }
+
+  // For worker/agent nodes without local kubectl: apply via the server
+  if (ctx.config.k3sServerUrl) {
+    // The CLI has SSH access to the server — use sshExec from there
+    const serverHost = new URL(ctx.config.k3sServerUrl).hostname;
+    try {
+      const remoteResult = await remoteSshExec(
+        serverHost, "root",
+        `k3s kubectl annotate node "${nodeName}" "node.longhorn.io/default-disks-config=${annotation}" --overwrite`,
+        { ...(ctx.ssh.keyPath ? { keyPath: ctx.ssh.keyPath } : {}), timeoutMs: 15_000 },
+      );
+      if (remoteResult.stdout.includes("annotated") || remoteResult.stdout.includes("unchanged")) {
+        return { success: true, changed: true, message: `Longhorn disk annotation applied to ${nodeName} (via server)` };
+      }
+    } catch {
+      // Fall through to manual instruction
+    }
+  }
+
+  return { success: true, changed: false, message: "Longhorn disk label set (annotation requires server kubectl)" };
+};

bastion/src/modules/modules/k3s/tests/install.test.ts

@@ -71,9 +71,14 @@ describe("k3s install script — server role", () => {
     expect(script).toContain("enable-admission-plugins=NodeRestriction,PodSecurity");
   });
 
-  it("configures audit logging", () => {
+  it("configures audit logging via journald (stdout)", () => {
-    expect(script).toContain("audit-log-path=/var/log/kubernetes/audit.log");
+    expect(script).toContain("audit-log-path=-");
-    expect(script).toContain("audit-log-maxage=30");
+    // file-based fields and the now-obsolete log directory must be gone
+    expect(script).not.toContain("/var/log/kubernetes/audit.log");
+    expect(script).not.toContain("audit-log-maxage");
+    expect(script).not.toContain("audit-log-maxbackup");
+    expect(script).not.toContain("audit-log-maxsize");
+    expect(script).not.toContain("mkdir -p /var/log/kubernetes");
   });
 
   it("cleans stale flannel vxlan before Cilium install", () => {

bastion/src/modules/modules/k3s/tests/operations.test.ts

@@ -348,3 +348,143 @@ describe("applyPodSecurityStandards", () => {
     expectCommand(ctx.ssh, "pod-security.kubernetes.io/audit=restricted");
   });
 });
+
+// --- Audit Logging Decommission (file-based → journald) ---
+
+import { configureLogRotation } from "../src/operations/log-rotation.js";
+import { configureJournaldLimits } from "../src/operations/journald-limits.js";
+
+describe("configureLogRotation (decommission file-based audit logs)", () => {
+  it("removes the legacy logrotate rule and reaps obsolete audit files", async () => {
+    const ctx = mockCtx();
+    ctx.ssh.exec.mockResolvedValueOnce(stdout("present")); // probe: legacy artifacts exist
+    ctx.ssh.exec.mockResolvedValue(OK);
+
+    const result = await configureLogRotation(ctx);
+
+    expect(result.success).toBe(true);
+    expect(result.changed).toBe(true);
+    expectCommand(ctx.ssh, "rm -f /etc/logrotate.d/k3s");
+    expectCommand(ctx.ssh, /find \/var\/log\/kubernetes.*audit.*-delete/);
+    expectCommand(ctx.ssh, "rmdir /var/log/kubernetes");
+  });
+
+  it("is a no-op when nothing legacy is present", async () => {
+    const ctx = mockCtx();
+    ctx.ssh.exec.mockResolvedValueOnce(stdout("absent"));
+    ctx.ssh.exec.mockResolvedValue(OK);
+
+    const result = await configureLogRotation(ctx);
+    expect(result.success).toBe(true);
+    expect(result.changed).toBe(false);
+  });
+});
+
+describe("configureJournaldLimits", () => {
+  it("writes a 2 GB SystemMaxUse drop-in and reloads journald when changed", async () => {
+    const ctx = mockCtx();
+    ctx.ssh.exec.mockResolvedValueOnce(stdout("__LABCTL_NOT_FOUND__")); // no existing drop-in
+    ctx.ssh.exec.mockResolvedValue(OK);
+
+    const result = await configureJournaldLimits(ctx);
+
+    expect(result.success).toBe(true);
+    expect(result.changed).toBe(true);
+    const writeCall = ctx.ssh.exec.mock.calls.find((c) => {
+      const cmd = c[0] as string;
+      return cmd.includes("10-k3s-audit-cap.conf") && cmd.includes("LABCTL_EOF");
+    });
+    expect(writeCall).toBeTruthy();
+    const written = writeCall?.[0] as string;
+    expect(written).toContain("SystemMaxUse=2G");
+    expect(written).toContain("SystemKeepFree=1G");
+    expectCommand(ctx.ssh, "systemctl restart systemd-journald");
+  });
+
+  it("does not restart journald when the drop-in is already correct", async () => {
+    const ctx = mockCtx();
+    const existing =
+      "[Journal]\nSystemMaxUse=2G\nSystemKeepFree=1G\nSystemMaxFileSize=200M\n";
+    ctx.ssh.exec.mockResolvedValueOnce(stdout(existing));
+    ctx.ssh.exec.mockResolvedValue(OK);
+
+    const result = await configureJournaldLimits(ctx);
+
+    expect(result.success).toBe(true);
+    expect(result.changed).toBe(false);
+    expectNoCommand(ctx.ssh, "systemctl restart systemd-journald");
+  });
+});
+
+// --- Etcd Recovery ---
+
+import { recoverEtcdMember } from "../src/operations/etcd-recover.js";
+import { mockSsh } from "./helpers.js";
+
+describe("recoverEtcdMember", () => {
+  it("refuses to operate when cluster is below 3 members (quorum risk)", async () => {
+    const broken = mockSsh();
+    const peer = mockSsh();
+    peer.exec.mockResolvedValueOnce(stdout("/usr/bin/etcdctl")); // etcdctl present
+    peer.exec.mockResolvedValueOnce(stdout(
+      "111, started, host-a-aaa, https://10.0.0.1:2380, https://10.0.0.1:2379, false\n" +
+      "222, started, host-b-bbb, https://10.0.0.2:2380, https://10.0.0.2:2379, false",
+    ));
+
+    const result = await recoverEtcdMember({ broken, peer, brokenHostname: "host-b" });
+
+    expect(result.success).toBe(false);
+    expect(result.message).toMatch(/quorum/i);
+    // Critically: must NOT have stopped k3s or removed anything
+    expect(broken.exec).not.toHaveBeenCalledWith(expect.stringContaining("systemctl stop k3s"), expect.anything());
+  });
+
+  it("performs full procedure when quorum is preserved", async () => {
+    const broken = mockSsh();
+    const peer = mockSsh();
+    // ensureEtcdctl: present
+    peer.exec.mockResolvedValueOnce(stdout("/usr/bin/etcdctl"));
+    // member list (3 members, target = host-b)
+    peer.exec.mockResolvedValueOnce(stdout(
+      "111, started, host-a-aaa, https://10.0.0.1:2380, https://10.0.0.1:2379, false\n" +
+      "222, started, host-b-bbb, https://10.0.0.2:2380, https://10.0.0.2:2379, false\n" +
+      "333, started, host-c-ccc, https://10.0.0.3:2380, https://10.0.0.3:2379, false",
+    ));
+    // member remove
+    peer.exec.mockResolvedValueOnce(stdout("Member 222 removed"));
+    // post-rejoin member list — new id 444 for host-b
+    peer.exec.mockResolvedValueOnce(stdout(
+      "111, started, host-a-aaa, https://10.0.0.1:2380, https://10.0.0.1:2379, false\n" +
+      "333, started, host-c-ccc, https://10.0.0.3:2380, https://10.0.0.3:2379, false\n" +
+      "444, started, host-b-zzz, https://10.0.0.2:2380, https://10.0.0.2:2379, false",
+    ));
+
+    const result = await recoverEtcdMember({ broken, peer, brokenHostname: "host-b" });
+
+    expect(result.success).toBe(true);
+    expect(result.removedMemberId).toBe("222");
+    expect(result.newMemberId).toBe("444");
+    expectCommand(broken,"systemctl stop k3s");
+    expectCommand(peer,"member remove 222");
+    expectCommand(broken,/db\.corrupt-/);
+    expectCommand(broken,/rm -rf .*\/server\/tls/);
+    expectCommand(broken,"systemctl start k3s");
+  });
+
+  it("fails clearly when no member matches the broken hostname", async () => {
+    const broken = mockSsh();
+    const peer = mockSsh();
+    peer.exec.mockResolvedValueOnce(stdout("/usr/bin/etcdctl"));
+    peer.exec.mockResolvedValueOnce(stdout(
+      "111, started, host-a-aaa, https://10.0.0.1:2380, https://10.0.0.1:2379, false\n" +
+      "222, started, host-b-bbb, https://10.0.0.2:2380, https://10.0.0.2:2379, false\n" +
+      "333, started, host-c-ccc, https://10.0.0.3:2380, https://10.0.0.3:2379, false",
+    ));
+
+    const result = await recoverEtcdMember({ broken, peer, brokenHostname: "host-d" });
+
+    expect(result.success).toBe(false);
+    expect(result.message).toMatch(/No etcd member found/);
+    expect(broken.exec).not.toHaveBeenCalledWith(expect.stringContaining("systemctl stop k3s"), expect.anything());
+  });
+});

bastion/src/shared/src/protocol/index.ts

@@ -113,6 +113,7 @@ export type LabdBastionMessage =
   | { type: "command-role-update"; requestId: string; mac: string; role: string }
   | { type: "command-debug"; requestId: string; mac: string; pxeBoot?: boolean }
   | { type: "command-register"; requestId: string; mac: string; hostname: string; role: string; ip: string }
+  | { type: "command-discover"; requestId: string; mac: string; product?: string; board?: string; serial?: string; manufacturer?: string; cpu_model?: string; cpu_cores?: number; memory_gb?: number; arch?: string; disks?: Array<{ name: string; size_gb: number; model: string }>; nics?: Array<{ name: string; mac: string; state: string }> }
   | { type: "server-shutdown"; reconnectAfter: number };
 
 export type BastionMessageType = BastionMessage["type"];
@@ -127,7 +128,7 @@ const BASTION_MESSAGE_TYPES = new Set<string>([
 
 const LABD_BASTION_MESSAGE_TYPES = new Set<string>([
   "bastion-enrolled", "bastion-heartbeat-ack", "command-install",
-  "command-forget", "command-role-update", "command-debug", "command-register", "server-shutdown",
+  "command-forget", "command-role-update", "command-debug", "command-register", "command-discover", "server-shutdown",
 ]);
 
 export function isBastionMessage(msg: unknown): msg is BastionMessage {

bastion/src/shared/src/types/state.ts

@@ -96,6 +96,13 @@ export interface InstalledInfo {
   ip: string;
   installed_at: string;
   bastionId?: string;  // set when aggregated through labd
+  // Hardware info (copied from discovered on install completion)
+  product?: string;
+  manufacturer?: string;
+  cpu_model?: string;
+  cpu_cores?: number;
+  memory_gb?: number;
+  arch?: string;
 }
 
 export interface DebugConfig {

bastion/tests/integration/asahi-firstboot.test.ts

@@ -0,0 +1,355 @@
+// Integration test: Asahi first-boot LVM setup.
+//
+// Tests the first-boot script that creates the standard lab LVM layout
+// on a separate data disk — simulating the Asahi provisioning flow where
+// the root partition is pre-installed and a data partition is left for LVM.
+//
+// Uses a Fedora cloud VM with two disks:
+//   disk0: 20GB root (Fedora cloud image)
+//   disk1: 200GB empty (simulates the Asahi "Data" partition)
+//
+// The firstboot script should detect disk1, create labvg + LVs, mount them.
+// Then we test reprovision: wipe marker, re-run, verify existing VG reused.
+//
+// Prerequisites: libvirt, virsh, virt-install, qemu, sudo access, lvm2
+// Run: sudo pnpm run test:integration:asahi
+
+import { describe, it, expect, beforeAll, afterAll } from "vitest";
+import { readFileSync, existsSync } from "node:fs";
+import { execSync } from "node:child_process";
+import { join } from "node:path";
+import { homedir } from "node:os";
+import { destroyVm, waitForVmIp, waitForSsh, log, ensureCloudImage, createCloudInitIso } from "./helpers/libvirt.js";
+import { ensureTestNetwork, TEST_NETWORK_NAME } from "./helpers/network.js";
+import { sshExec, sshRun } from "./helpers/ssh.js";
+import { renderFirstbootScript } from "../../src/bastion/src/templates/asahi-firstboot.sh.js";
+
+const VM_NAME = "lab-asahi-firstboot-test";
+const VM_MEMORY = 4096;
+const VM_VCPUS = 2;
+const VM_ROOT_DISK_GB = 20;
+const VM_DATA_DISK_GB = 200;    // Simulates the Asahi "Data" partition
+const SSH_USER = "fedora";
+const IMAGE_DIR = "/var/lib/libvirt/images";
+const IS_ROOT = process.getuid?.() === 0;
+
+const FEDORA_CLOUD_IMAGE = "https://download.fedoraproject.org/pub/fedora/linux/releases/43/Cloud/x86_64/images/Fedora-Cloud-Base-Generic-43-1.6.x86_64.qcow2";
+
+function run(cmd: string, opts?: { timeout?: number }): string {
+  const full = IS_ROOT ? cmd : `sudo ${cmd}`;
+  return execSync(full, { encoding: "utf-8", stdio: "pipe", timeout: opts?.timeout ?? 60_000 });
+}
+
+function findSshKey(): { pubKey: string; keyPath: string } {
+  const homes = [homedir()];
+  const sudoUser = process.env["SUDO_USER"];
+  if (sudoUser) homes.push(join("/home", sudoUser));
+  if (process.env["SSH_KEY_PATH"]) {
+    const keyPath = process.env["SSH_KEY_PATH"];
+    const pubPath = `${keyPath}.pub`;
+    if (existsSync(keyPath) && existsSync(pubPath)) {
+      return { pubKey: readFileSync(pubPath, "utf-8").trim(), keyPath };
+    }
+  }
+  for (const home of homes) {
+    for (const name of ["id_ed25519", "id_ecdsa", "id_rsa"]) {
+      const keyPath = join(home, ".ssh", name);
+      const pubPath = `${keyPath}.pub`;
+      if (existsSync(keyPath) && existsSync(pubPath)) {
+        return { pubKey: readFileSync(pubPath, "utf-8").trim(), keyPath };
+      }
+    }
+  }
+  throw new Error("No SSH key found");
+}
+
+/** Create a VM with two disks: root (cloud image) + empty data disk. */
+function createTwoDiskVm(config: {
+  name: string;
+  memory: number;
+  vcpus: number;
+  rootDiskGb: number;
+  dataDiskGb: number;
+  network: string;
+  cloudImageUrl: string;
+  sshPubKey: string;
+}): void {
+  destroyVm(config.name);
+
+  log(`Creating two-disk VM: ${config.name} (root=${config.rootDiskGb}GB, data=${config.dataDiskGb}GB)`);
+
+  const baseImage = ensureCloudImage(config.cloudImageUrl, `${config.name}-base`);
+  const rootDiskPath = join(IMAGE_DIR, `${config.name}.qcow2`);
+  const dataDiskPath = join(IMAGE_DIR, `${config.name}-data.qcow2`);
+
+  // Root disk from cloud image
+  run(`cp "${baseImage}" "${rootDiskPath}"`);
+  run(`qemu-img resize "${rootDiskPath}" ${config.rootDiskGb}G`);
+
+  // Empty data disk
+  run(`qemu-img create -f qcow2 "${dataDiskPath}" ${config.dataDiskGb}G`);
+
+  // Cloud-init with LVM tools
+  const cloudInitIso = createCloudInitIso(config.name, {
+    name: config.name,
+    memory: config.memory,
+    vcpus: config.vcpus,
+    diskSize: config.rootDiskGb,
+    network: config.network,
+    cloudImageUrl: config.cloudImageUrl,
+    sshPubKey: config.sshPubKey,
+    userData: `#cloud-config
+hostname: ${config.name}
+manage_etc_hosts: true
+users:
+  - default
+  - name: fedora
+    sudo: ALL=(ALL) NOPASSWD:ALL
+    shell: /bin/bash
+    ssh_authorized_keys:
+      - ${config.sshPubKey}
+ssh_pwauth: false
+package_update: false
+packages:
+  - lvm2
+  - xfsprogs
+`,
+  });
+
+  const virtInstallArgs = [
+    "virt-install",
+    `--name=${config.name}`,
+    `--memory=${config.memory}`,
+    `--vcpus=${config.vcpus}`,
+    `--disk=path=${rootDiskPath},format=qcow2`,
+    `--disk=path=${dataDiskPath},format=qcow2`,  // Second disk for LVM
+    `--disk=path=${cloudInitIso},device=cdrom`,
+    `--network=network=${config.network},model=virtio`,
+    "--os-variant=generic",
+    "--import",
+    "--noautoconsole",
+    "--wait=0",
+  ];
+
+  run(virtInstallArgs.join(" "));
+  log(`Two-disk VM ${config.name} created`);
+}
+
+describe("asahi firstboot LVM integration", () => {
+  let vmIp: string;
+  let sshKeyPath: string;
+  let sshPubKey: string;
+
+  beforeAll(async () => {
+    const keys = findSshKey();
+    sshKeyPath = keys.keyPath;
+    sshPubKey = keys.pubKey;
+
+    log("Setting up test network...");
+    ensureTestNetwork();
+
+    log("Creating two-disk VM...");
+    createTwoDiskVm({
+      name: VM_NAME,
+      memory: VM_MEMORY,
+      vcpus: VM_VCPUS,
+      rootDiskGb: VM_ROOT_DISK_GB,
+      dataDiskGb: VM_DATA_DISK_GB,
+      network: TEST_NETWORK_NAME,
+      cloudImageUrl: FEDORA_CLOUD_IMAGE,
+      sshPubKey,
+    });
+
+    log("Waiting for VM IP...");
+    vmIp = await waitForVmIp(VM_NAME, 120_000);
+
+    log("Waiting for SSH...");
+    await waitForSsh(vmIp, SSH_USER, 180_000, sshKeyPath);
+
+    log("Waiting for cloud-init to finish...");
+    await sshRun(vmIp, SSH_USER, "sudo cloud-init status --wait 2>/dev/null || sleep 30", "cloud-init", { keyPath: sshKeyPath });
+
+    // Verify second disk exists
+    const disks = sshExec(vmIp, SSH_USER, "lsblk -d -n -o NAME,SIZE", { keyPath: sshKeyPath });
+    log(`Disks:\n${disks.stdout}`);
+  }, 300_000);
+
+  afterAll(async () => {
+    log("Cleaning up VM...");
+    destroyVm(VM_NAME);
+    // Also remove data disk
+    try { run(`rm -f "${join(IMAGE_DIR, `${VM_NAME}-data.qcow2`)}"`); } catch { /* ignore */ }
+  });
+
+  it("second disk is visible and unformatted", () => {
+    const result = sshExec(vmIp, SSH_USER, "lsblk -d -n -o NAME,SIZE,TYPE | grep disk", { keyPath: sshKeyPath });
+    const disks = result.stdout.trim().split("\n");
+    expect(disks.length).toBeGreaterThanOrEqual(2);
+
+    // Second disk (vdb) should exist
+    const vdb = sshExec(vmIp, SSH_USER, "sudo blkid /dev/vdb 2>/dev/null; echo exit=$?", { keyPath: sshKeyPath });
+    // Should have no filesystem (blkid returns nothing or non-zero)
+    expect(vdb.stdout).toContain("exit=2");
+  });
+
+  it("firstboot script creates LVM on data disk", async () => {
+    // Generate the firstboot script
+    const script = renderFirstbootScript({
+      hostname: "asahi-test",
+      role: "infra",
+      serverIp: "10.0.0.1",
+      httpPort: 8080,
+      sshKeys: [sshPubKey],
+      adminUser: "testadmin",
+      mac: "52:54:00:aa:bb:cc",
+    });
+
+    // Upload and run
+    log("Uploading firstboot script...");
+    await sshRun(vmIp, SSH_USER,
+      `cat > /tmp/firstboot.sh << 'SCRIPT_EOF'\n${script}\nSCRIPT_EOF\nchmod +x /tmp/firstboot.sh`,
+      "upload script", { keyPath: sshKeyPath });
+
+    log("Running firstboot script...");
+    const result = await sshRun(vmIp, SSH_USER,
+      "sudo /tmp/firstboot.sh 2>&1",
+      "firstboot", { keyPath: sshKeyPath, timeout: 120_000 });
+
+    expect(result).toBe(0);
+  }, 180_000);
+
+  it("SSH still works after firstboot script", () => {
+    const result = sshExec(vmIp, SSH_USER, "echo hello", { keyPath: sshKeyPath });
+    if (result.stdout.trim() !== "hello") {
+      log(`SSH debug: exitCode=${result.exitCode} stdout='${result.stdout}' stderr='${result.stderr}'`);
+    }
+    expect(result.stdout.trim()).toBe("hello");
+  });
+
+  it("volume group labvg exists", () => {
+    const result = sshExec(vmIp, SSH_USER, "sudo vgs labvg --noheadings -o vg_name", { keyPath: sshKeyPath });
+    expect(result.stdout.trim()).toBe("labvg");
+  });
+
+  it("all expected logical volumes exist", () => {
+    const result = sshExec(vmIp, SSH_USER,
+      "sudo lvs labvg --noheadings -o lv_name --sort lv_name",
+      { keyPath: sshKeyPath });
+    const lvs = result.stdout.trim().split("\n").map(l => l.trim()).sort();
+    expect(lvs).toContain("home");
+    expect(lvs).toContain("longhorn");
+    expect(lvs).toContain("rancher");   // infra role
+    expect(lvs).toContain("srv");
+    expect(lvs).toContain("swap");
+    expect(lvs).toContain("var");
+    expect(lvs).toContain("varlog");
+  });
+
+  it("LV sizes match kickstart layout", () => {
+    const result = sshExec(vmIp, SSH_USER,
+      "sudo lvs labvg --noheadings -o lv_name,lv_size --units m --nosuffix",
+      { keyPath: sshKeyPath });
+    const lvMap = new Map<string, number>();
+    for (const line of result.stdout.trim().split("\n")) {
+      const [name, size] = line.trim().split(/\s+/);
+      if (name && size) lvMap.set(name, Math.round(parseFloat(size)));
+    }
+
+    expect(lvMap.get("swap")).toBe(27648);
+    expect(lvMap.get("var")).toBe(102400);
+    expect(lvMap.get("varlog")).toBe(10240);
+    expect(lvMap.get("home")).toBe(10240);
+    expect(lvMap.get("srv")).toBe(20480);
+    expect(lvMap.get("rancher")).toBe(20480);
+    // longhorn gets remaining — should be at least 5GB (200GB disk - ~191GB used)
+    expect(lvMap.get("longhorn")).toBeGreaterThan(5000);
+  });
+
+  it("non-var volumes are mounted with XFS", () => {
+    const mounts = sshExec(vmIp, SSH_USER, "mount | grep labvg", { keyPath: sshKeyPath });
+    // /var and /var/log deferred to next reboot (can't migrate live)
+    expect(mounts.stdout).toContain("/home ");
+    expect(mounts.stdout).toContain("/srv ");
+    expect(mounts.stdout).toContain("/var/lib/rancher ");
+    expect(mounts.stdout).toContain("/var/lib/longhorn ");
+    expect(mounts.stdout).toContain("xfs");
+  });
+
+  it("swap is active", () => {
+    const result = sshExec(vmIp, SSH_USER, "swapon --show --noheadings", { keyPath: sshKeyPath });
+    // swapon may show /dev/dm-X or /dev/labvg/swap
+    expect(result.stdout.length).toBeGreaterThan(0);
+  });
+
+  it("fstab has LVM entries", () => {
+    const result = sshExec(vmIp, SSH_USER, "grep labvg /etc/fstab", { keyPath: sshKeyPath });
+    const lines = result.stdout.trim().split("\n");
+    expect(lines.length).toBeGreaterThanOrEqual(7); // swap + var + varlog + home + srv + rancher + longhorn
+  });
+
+  it("hostname was set", () => {
+    const result = sshExec(vmIp, SSH_USER, "hostname", { keyPath: sshKeyPath });
+    expect(result.stdout.trim()).toBe("asahi-test");
+  });
+
+  it("admin user was created with sudo", () => {
+    const result = sshExec(vmIp, SSH_USER, "sudo id testadmin", { keyPath: sshKeyPath });
+    expect(result.stdout).toContain("testadmin");
+    expect(result.stdout).toContain("wheel");
+  });
+
+  it("provisioning metadata file exists", () => {
+    const result = sshExec(vmIp, SSH_USER, "cat /etc/lab-provisioned", { keyPath: sshKeyPath });
+    expect(result.stdout).toContain("hostname=asahi-test");
+    expect(result.stdout).toContain("role=infra");
+    expect(result.stdout).toContain("method=asahi-firstboot");
+  });
+
+  it("marker file prevents re-run", () => {
+    const result = sshExec(vmIp, SSH_USER, "test -f /etc/lab-lvm-setup-done && echo yes", { keyPath: sshKeyPath });
+    expect(result.stdout.trim()).toBe("yes");
+  });
+
+  // ── Reprovision test ──────────────────────────────────────────────
+
+  it("reprovision: detects existing labvg and re-mounts", async () => {
+    // Write a test file to a preserved LV
+    await sshRun(vmIp, SSH_USER,
+      "echo 'precious-data' | sudo tee /var/lib/rancher/test-preserve.txt",
+      "write test data", { keyPath: sshKeyPath });
+
+    // Remove marker to simulate fresh boot after reinstall
+    await sshRun(vmIp, SSH_USER, "sudo rm /etc/lab-lvm-setup-done", "remove marker", { keyPath: sshKeyPath });
+
+    // Unmount everything (simulate reinstall wiping root)
+    await sshRun(vmIp, SSH_USER, `
+      sudo umount /var/lib/longhorn 2>/dev/null || true
+      sudo umount /var/lib/rancher 2>/dev/null || true
+      sudo umount /srv 2>/dev/null || true
+      sudo umount /home 2>/dev/null || true
+      sudo umount /var/log 2>/dev/null || true
+      # Don't unmount /var — it's in use
+      sudo swapoff /dev/labvg/swap 2>/dev/null || true
+      sudo sed -i '/labvg/d' /etc/fstab
+    `, "unmount LVs", { keyPath: sshKeyPath });
+
+    // Re-run firstboot script — should detect existing VG
+    log("Re-running firstboot (reprovision)...");
+    const result = await sshRun(vmIp, SSH_USER,
+      "sudo /tmp/firstboot.sh 2>&1",
+      "firstboot reprovision", { keyPath: sshKeyPath });
+    expect(result).toBe(0);
+
+    // Verify data was preserved
+    const data = sshExec(vmIp, SSH_USER, "cat /var/lib/rancher/test-preserve.txt", { keyPath: sshKeyPath });
+    expect(data.stdout.trim()).toBe("precious-data");
+
+    // Verify marker was re-created
+    const marker = sshExec(vmIp, SSH_USER, "test -f /etc/lab-lvm-setup-done && echo yes", { keyPath: sshKeyPath });
+    expect(marker.stdout.trim()).toBe("yes");
+
+    // Verify fstab was re-populated
+    const fstab = sshExec(vmIp, SSH_USER, "grep labvg /etc/fstab", { keyPath: sshKeyPath });
+    expect(fstab.stdout).toContain("/var/lib/rancher");
+  }, 60_000);
+});

bastion/tests/integration/asahi-validate.test.ts

@@ -0,0 +1,353 @@
+// Validation tests for Asahi provisioning artifacts.
+//
+// Tests that can run WITHOUT Apple Silicon hardware:
+//   1. Shellcheck the generated firstboot script
+//   2. Verify the built rootfs ZIP structure
+//   3. Mount the rootfs and verify injected files
+//   4. Validate installer_data.json against the Asahi installer's Python parser
+//   5. Verify partition layout arithmetic
+//
+// Prerequisites:
+//   - Run scripts/build-asahi-rootfs.sh first (creates asahi-repo/)
+//   - shellcheck installed (dnf install ShellCheck)
+//   - python3 installed
+//   - root for loop mount (sudo)
+//
+// Run: sudo pnpm run test:integration:asahi-validate
+
+import { describe, it, expect, beforeAll, afterAll } from "vitest";
+import { existsSync, lstatSync, readFileSync, writeFileSync, mkdirSync, rmSync } from "node:fs";
+import { execSync, spawnSync } from "node:child_process";
+import { join } from "node:path";
+import { tmpdir } from "node:os";
+import { renderFirstbootScript } from "../../src/bastion/src/templates/asahi-firstboot.sh.js";
+
+const PROJECT_ROOT = join(import.meta.dirname, "..", "..");
+const ASAHI_REPO = join(PROJECT_ROOT, "asahi-repo");
+const ASAHI_CACHE = join(PROJECT_ROOT, ".asahi-cache");
+const IS_ROOT = process.getuid?.() === 0;
+
+function run(cmd: string, opts?: { timeout?: number }): string {
+  const full = IS_ROOT ? cmd : `sudo ${cmd}`;
+  return execSync(full, { encoding: "utf-8", stdio: "pipe", timeout: opts?.timeout ?? 60_000 });
+}
+
+function hasBuiltArtifacts(): boolean {
+  return existsSync(join(ASAHI_REPO, "fedora-asahi-lab.zip")) &&
+         existsSync(join(ASAHI_REPO, "installer_data.json"));
+}
+
+describe("asahi script validation", () => {
+  it("firstboot script passes shellcheck", () => {
+    const script = renderFirstbootScript({
+      hostname: "test-node",
+      role: "infra",
+      serverIp: "10.0.0.1",
+      httpPort: 8080,
+      sshKeys: ["ssh-ed25519 AAAA... user@host"],
+      adminUser: "testadmin",
+      mac: "aa:bb:cc:dd:ee:ff",
+    });
+
+    const tmpFile = join(tmpdir(), `asahi-shellcheck-${Date.now()}.sh`);
+    writeFileSync(tmpFile, script);
+
+    try {
+      const result = spawnSync("shellcheck", [
+        "-s", "bash",
+        "-e", "SC2086,SC2164",  // allow unquoted variables (intentional in some LVM commands)
+        tmpFile,
+      ], { encoding: "utf-8", stdio: "pipe", timeout: 30_000 });
+
+      if (result.status !== 0) {
+        console.log("Shellcheck warnings/errors:");
+        console.log(result.stdout);
+      }
+      // Allow warnings (exit 1 for warnings), fail on errors (exit 2+)
+      expect(result.status).toBeLessThan(2);
+    } finally {
+      try { rmSync(tmpFile); } catch { /* ignore */ }
+    }
+  });
+
+  it("firstboot script for worker role passes shellcheck", () => {
+    const script = renderFirstbootScript({
+      hostname: "worker-node",
+      role: "worker",
+      serverIp: "10.0.0.1",
+      httpPort: 8080,
+      sshKeys: [],
+      adminUser: "michal",
+      mac: "00:11:22:33:44:55",
+    });
+
+    const tmpFile = join(tmpdir(), `asahi-shellcheck-worker-${Date.now()}.sh`);
+    writeFileSync(tmpFile, script);
+
+    try {
+      const result = spawnSync("shellcheck", ["-s", "bash", "-e", "SC2086,SC2164", tmpFile],
+        { encoding: "utf-8", stdio: "pipe", timeout: 30_000 });
+      if (result.status !== 0) console.log(result.stdout);
+      expect(result.status).toBeLessThan(2);
+    } finally {
+      try { rmSync(tmpFile); } catch { /* ignore */ }
+    }
+  });
+
+  it("firstboot script for vanilla role passes shellcheck", () => {
+    const script = renderFirstbootScript({
+      hostname: "vanilla-node",
+      role: "vanilla",
+      serverIp: "10.0.0.1",
+      httpPort: 8080,
+      sshKeys: ["ssh-rsa AAAA... user@host"],
+      adminUser: "admin",
+      mac: "ff:ee:dd:cc:bb:aa",
+    });
+
+    const tmpFile = join(tmpdir(), `asahi-shellcheck-vanilla-${Date.now()}.sh`);
+    writeFileSync(tmpFile, script);
+
+    try {
+      const result = spawnSync("shellcheck", ["-s", "bash", "-e", "SC2086,SC2164", tmpFile],
+        { encoding: "utf-8", stdio: "pipe", timeout: 30_000 });
+      if (result.status !== 0) console.log(result.stdout);
+      expect(result.status).toBeLessThan(2);
+    } finally {
+      try { rmSync(tmpFile); } catch { /* ignore */ }
+    }
+  });
+});
+
+describe("asahi installer_data.json validation", () => {
+  let installerData: Record<string, unknown>;
+
+  beforeAll(() => {
+    if (!hasBuiltArtifacts()) {
+      throw new Error("Run scripts/build-asahi-rootfs.sh first to generate artifacts");
+    }
+    installerData = JSON.parse(readFileSync(join(ASAHI_REPO, "installer_data.json"), "utf-8"));
+  });
+
+  it("has os_list with one entry", () => {
+    const osList = installerData["os_list"] as unknown[];
+    expect(osList).toBeInstanceOf(Array);
+    expect(osList.length).toBe(1);
+  });
+
+  it("has required top-level fields", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    expect(os["name"]).toBeDefined();
+    expect(os["default_os_name"]).toBeDefined();
+    expect(os["boot_object"]).toBeDefined();
+    expect(os["next_object"]).toBeDefined();
+    expect(os["package"]).toBe("fedora-asahi-lab.zip");
+    expect(os["supported_fw"]).toBeInstanceOf(Array);
+    expect((os["supported_fw"] as string[]).length).toBeGreaterThan(0);
+  });
+
+  it("has 4 partitions (EFI + Boot + Root + Data)", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const partitions = os["partitions"] as Record<string, unknown>[];
+    expect(partitions).toHaveLength(4);
+    expect(partitions[0]!["name"]).toBe("EFI");
+    expect(partitions[1]!["name"]).toBe("Boot");
+    expect(partitions[2]!["name"]).toBe("Root");
+    expect(partitions[3]!["name"]).toBe("Data");
+  });
+
+  it("EFI partition has correct format", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const efi = (os["partitions"] as Record<string, unknown>[])[0]!;
+    expect(efi["type"]).toBe("EFI");
+    expect(efi["format"]).toBe("fat");
+    expect(efi["copy_firmware"]).toBe(true);
+    // Size should be ~500MB in bytes
+    const size = parseInt(String(efi["size"]).replace("B", ""), 10);
+    expect(size).toBeGreaterThanOrEqual(500 * 1024 * 1024);
+  });
+
+  it("Boot partition references boot.img", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const boot = (os["partitions"] as Record<string, unknown>[])[1]!;
+    expect(boot["type"]).toBe("Linux");
+    expect(boot["image"]).toBe("boot.img");
+  });
+
+  it("Root partition does NOT expand", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const root = (os["partitions"] as Record<string, unknown>[])[2]!;
+    expect(root["type"]).toBe("Linux");
+    expect(root["image"]).toBe("root.img");
+    expect(root["expand"]).toBe(false);
+  });
+
+  it("Data partition expands for LVM", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const data = (os["partitions"] as Record<string, unknown>[])[3]!;
+    expect(data["type"]).toBe("Linux");
+    expect(data["expand"]).toBe(true);
+    expect(data["image"]).toBeUndefined(); // No image — empty partition for LVM
+  });
+
+  it("partition sizes use bytes format (NB suffix)", () => {
+    const os = (installerData["os_list"] as Record<string, unknown>[])[0]!;
+    const partitions = os["partitions"] as Record<string, unknown>[];
+    for (const p of partitions) {
+      const size = String(p["size"]);
+      expect(size).toMatch(/^\d+B$/);
+    }
+  });
+
+  it("validates against Asahi installer Python parser", () => {
+    // Download the Asahi installer and run its validation logic on our config
+    const validation = spawnSync("python3", ["-c", `
+import json, sys
+
+with open("${join(ASAHI_REPO, "installer_data.json")}") as f:
+    data = json.load(f)
+
+errors = []
+os_list = data.get("os_list", [])
+if not os_list:
+    errors.append("Empty os_list")
+
+for os_entry in os_list:
+    required = ["name", "default_os_name", "boot_object", "next_object", "package", "supported_fw", "partitions"]
+    for field in required:
+        if field not in os_entry:
+            errors.append(f"Missing field: {field}")
+
+    partitions = os_entry.get("partitions", [])
+    if not partitions:
+        errors.append("No partitions defined")
+
+    has_efi = False
+    has_root_image = False
+    expand_count = 0
+    for p in partitions:
+        if "name" not in p or "type" not in p or "size" not in p:
+            errors.append(f"Partition missing name/type/size: {p}")
+        if p.get("type") == "EFI":
+            has_efi = True
+            if p.get("format") != "fat":
+                errors.append("EFI partition must be FAT format")
+        if p.get("image"):
+            has_root_image = True
+        if p.get("expand"):
+            expand_count += 1
+        # Validate size format
+        size_str = str(p.get("size", ""))
+        if not size_str.endswith("B") or not size_str[:-1].isdigit():
+            errors.append(f"Invalid size format: {size_str} (expected NB)")
+
+    if not has_efi:
+        errors.append("No EFI partition found")
+    if not has_root_image:
+        errors.append("No partition with root image found")
+    if expand_count > 1:
+        errors.append(f"Multiple expanding partitions ({expand_count}) — only one should expand")
+
+    # Verify supported_fw is a list of strings
+    fw = os_entry.get("supported_fw", [])
+    if not isinstance(fw, list) or not all(isinstance(v, str) for v in fw):
+        errors.append("supported_fw must be a list of strings")
+
+if errors:
+    print("ERRORS:")
+    for e in errors:
+        print(f"  - {e}")
+    sys.exit(1)
+else:
+    print("OK: installer_data.json is valid")
+`], { encoding: "utf-8", stdio: "pipe", timeout: 10_000 });
+
+    if (validation.status !== 0) {
+      console.log(validation.stdout);
+      console.log(validation.stderr);
+    }
+    expect(validation.stdout).toContain("OK");
+    expect(validation.status).toBe(0);
+  });
+});
+
+describe("asahi rootfs ZIP validation", () => {
+  beforeAll(() => {
+    if (!hasBuiltArtifacts()) {
+      throw new Error("Run scripts/build-asahi-rootfs.sh first to generate artifacts");
+    }
+  });
+
+  it("ZIP contains required files", () => {
+    const result = spawnSync("unzip", ["-l", join(ASAHI_REPO, "fedora-asahi-lab.zip")],
+      { encoding: "utf-8", stdio: "pipe", timeout: 10_000 });
+    expect(result.stdout).toContain("boot.img");
+    expect(result.stdout).toContain("root.img");
+    expect(result.stdout).toContain("esp/");
+  });
+
+  it("boot.img is ~1GB", () => {
+    const result = spawnSync("unzip", ["-l", join(ASAHI_REPO, "fedora-asahi-lab.zip")],
+      { encoding: "utf-8", stdio: "pipe", timeout: 10_000 });
+    const bootLine = result.stdout.split("\n").find(l => l.includes("boot.img") && !l.includes("/"));
+    expect(bootLine).toBeDefined();
+    const size = parseInt(bootLine!.trim().split(/\s+/)[0]!, 10);
+    expect(size).toBeGreaterThan(500 * 1024 * 1024);  // > 500MB
+    expect(size).toBeLessThan(2 * 1024 * 1024 * 1024); // < 2GB
+  });
+
+  it("root.img is > 3GB", () => {
+    const result = spawnSync("unzip", ["-l", join(ASAHI_REPO, "fedora-asahi-lab.zip")],
+      { encoding: "utf-8", stdio: "pipe", timeout: 10_000 });
+    const rootLine = result.stdout.split("\n").find(l => l.includes("root.img"));
+    expect(rootLine).toBeDefined();
+    const size = parseInt(rootLine!.trim().split(/\s+/)[0]!, 10);
+    expect(size).toBeGreaterThan(3 * 1024 * 1024 * 1024); // > 3GB
+  });
+
+  it("rootfs contains lab-firstboot.sh", () => {
+    const mountDir = join(tmpdir(), `asahi-rootfs-check-${Date.now()}`);
+    const extractDir = join(tmpdir(), `asahi-rootfs-extract-${Date.now()}`);
+    mkdirSync(mountDir);
+    mkdirSync(extractDir);
+
+    try {
+      // Extract root.img from ZIP
+      run(`unzip -o -j "${join(ASAHI_REPO, "fedora-asahi-lab.zip")}" root.img -d "${extractDir}"`);
+
+      // Mount and check
+      run(`mount -o loop,ro "${join(extractDir, "root.img")}" "${mountDir}"`);
+
+      // Verify firstboot script
+      expect(existsSync(join(mountDir, "usr/local/bin/lab-firstboot.sh"))).toBe(true);
+      const script = readFileSync(join(mountDir, "usr/local/bin/lab-firstboot.sh"), "utf-8");
+      expect(script).toContain("#!/bin/bash");
+      expect(script).toContain("labvg");
+      expect(script).toContain("pvcreate");
+
+      // Verify systemd service
+      expect(existsSync(join(mountDir, "etc/systemd/system/lab-firstboot.service"))).toBe(true);
+      const service = readFileSync(join(mountDir, "etc/systemd/system/lab-firstboot.service"), "utf-8");
+      expect(service).toContain("lab-firstboot.sh");
+
+      // Verify service is enabled (symlink exists)
+      const symlinkPath = join(mountDir, "etc/systemd/system/multi-user.target.wants/lab-firstboot.service");
+      let symlinkExists = false;
+      try { lstatSync(symlinkPath); symlinkExists = true; } catch { /* not found */ }
+      expect(symlinkExists).toBe(true);
+
+      // Verify SSH keys
+      expect(existsSync(join(mountDir, "root/.ssh/authorized_keys"))).toBe(true);
+
+      // Verify lvm2 + xfsprogs are in the image
+      const hasLvm = existsSync(join(mountDir, "usr/bin/pvcreate")) || existsSync(join(mountDir, "usr/sbin/pvcreate"));
+      const hasXfs = existsSync(join(mountDir, "usr/bin/mkfs.xfs")) || existsSync(join(mountDir, "usr/sbin/mkfs.xfs"));
+      expect(hasLvm).toBe(true);
+      expect(hasXfs).toBe(true);
+    } finally {
+      run(`umount "${mountDir}" 2>/dev/null || true`);
+      rmSync(mountDir, { recursive: true, force: true });
+      rmSync(extractDir, { recursive: true, force: true });
+    }
+  }, 120_000);
+});