kubesolo-os/docs/design/kubesolo-os-design.md

KubeSolo OS — Bootable Immutable Kubernetes Distribution

Design Research: KubeSolo + Tiny Core Linux

---

1. Executive Summary

This document outlines the architecture for KubeSolo OS — an immutable, bootable Linux distribution purpose-built to run KubeSolo (Portainer's single-node Kubernetes distribution) with atomic updates. The design combines the minimal footprint of Tiny Core Linux with KubeSolo's single-binary K8s packaging to create an appliance-like Kubernetes node that boots directly into a production-ready cluster.

Target use cases: IoT/IIoT edge devices, single-node K8s appliances, air-gapped deployments, embedded systems, kiosk/POS systems, and resource-constrained hardware.

---

2. Component Analysis

2.1 KubeSolo

Source: https://github.com/portainer/kubesolo

KubeSolo is Portainer's production-ready, ultra-lightweight single-node Kubernetes distribution designed for edge and IoT scenarios.

Architecture highlights:

• Single binary — all K8s components bundled into one executable
• SQLite backend — uses Kine to replace etcd, eliminating cluster coordination overhead
• Bundled runtime — ships containerd, runc, CoreDNS, and CNI plugins
• No scheduler — replaced by a custom NodeSetter admission webhook (single-node, no scheduling decisions needed)
• Dual libc support — detects and supports both glibc and musl (Alpine) environments
• Offline-ready — designed for air-gapped deployments; all images can be preloaded
• Portainer Edge integration — optional remote management via --portainer-edge-* flags

Runtime requirements:

Requirement	Minimum	Recommended
RAM	512 MB	1 GB+
Kernel	3.10+ (legacy)	5.8+ (cgroup v2)
Storage	~500 MB (binary)	2 GB+ (with workloads)

Key kernel dependencies:

• cgroup v2 (kernel 5.8+) — CONFIG_CGROUP, CONFIG_CGROUP_CPUACCT, CONFIG_CGROUP_DEVICE, CONFIG_CGROUP_FREEZER, CONFIG_CGROUP_SCHED, CONFIG_CGROUP_PIDS, CONFIG_CGROUP_NET_CLASSID
• Namespaces — CONFIG_NAMESPACES, CONFIG_NET_NS, CONFIG_PID_NS, CONFIG_USER_NS, CONFIG_UTS_NS, CONFIG_IPC_NS
• Networking — CONFIG_BRIDGE, CONFIG_NETFILTER, CONFIG_VETH, CONFIG_VXLAN, CONFIG_IP_NF_IPTABLES, CONFIG_IP_NF_NAT
• Filesystem — CONFIG_OVERLAY_FS, CONFIG_SQUASHFS
• Modules required at runtime: br_netfilter, overlay, ip_tables, iptable_nat, iptable_filter

Installation & operation:

# Standard install
curl -sfL https://get.kubesolo.io | sudo sh -

# Kubeconfig location
/var/lib/kubesolo/pki/admin/admin.kubeconfig

# Key flags
--path /var/lib/kubesolo     # config directory
--apiserver-extra-sans        # additional TLS SANs
--local-storage true          # enable local-path provisioner
--portainer-edge-id           # Portainer Edge agent ID
--portainer-edge-key          # Portainer Edge agent key

2.2 Tiny Core Linux

Source: http://www.tinycorelinux.net

Tiny Core Linux is an ultra-minimal Linux distribution (11–17 MB) that runs entirely in RAM.

Architecture highlights:

• Micro Core — 11 MB: kernel + root filesystem + basic kernel modules (no GUI)
• RAM-resident — entire OS loaded into memory at boot; disk only needed for persistence
• SquashFS root — read-only compressed filesystem, inherently immutable
• Extension system — .tcz packages (SquashFS-compressed) mounted or copied at boot
• Three operational modes:
  1. Cloud/Default — pure RAM, nothing persists across reboots
  2. Mount mode — extensions stored in /tce directory, loop-mounted at boot
  3. Copy mode — extensions copied into RAM from persistent storage

Key concepts for this design:

• /tce directory on persistent storage holds extensions and configuration
• onboot.lst — list of extensions to auto-mount at boot
• filetool.sh + /opt/.filetool.lst — backup/restore mechanism for persistent files
• Boot codes control behavior: tce=, base, norestore, noswap, etc.
• Custom remastering: extract core.gz → modify → repack → create bootable image
• Frugal install: vmlinuz + core.gz + bootloader + /tce directory

Kernel: Ships modern Linux kernel (6.x series in v17.0), supports x86, x86_64, ARM.

---

3. Competitive Landscape — Existing Immutable K8s OSes

3.1 Comparison Matrix

Feature	Talos Linux	Bottlerocket	Flatcar Linux	Kairos	KubeSolo OS (proposed)
Footprint	~80 MB	~500 MB	~700 MB	Varies (base distro)	~50–80 MB
Immutability	Radical (12 binaries)	Strong (read-only root)	Moderate (read-only /usr)	Strong (overlayFS)	Strong (SquashFS root)
SSH access	None (API only)	Disabled (container shell)	Yes	Optional	Optional (extension)
Update model	A/B partitions	A/B partitions	A/B partitions (ChromeOS)	A/B partitions (OCI)	A/B partitions
K8s variants	Multi-node, HA	Multi-node (EKS)	Multi-node (any)	Multi-node (any)	Single-node only
Management	talosctl (mTLS API)	API (localhost)	Ignition + SSH	Cloud-init, K8s CRDs	API + cloud-init
Base OS	Custom (Go userland)	Custom (Bottlerocket)	Gentoo-derived	Any Linux (meta-distro)	Tiny Core Linux
Target	Cloud + Edge	AWS (primarily)	Cloud + Bare metal	Edge + Bare metal	Edge + IoT
Configuration	Machine config YAML	TOML settings	Ignition JSON	Cloud-init YAML	Cloud-init + boot codes

3.2 Key Lessons from Each

From Talos Linux:

• API-only management is powerful but aggressive — provide as optional mode
• 12-binary minimalism is aspirational; KubeSolo's single binary aligns well
• System extensions as SquashFS overlays in initramfs = directly applicable to Tiny Core's .tcz model
• A/B partition with GRUB fallback counter for automatic rollback

From Bottlerocket:

• Bootstrap containers for customization — useful pattern for pre-deploying workloads
• Host containers for privileged operations (debugging, admin access)
• Tightly coupled OS+K8s versions simplifies compatibility testing

From Flatcar Linux:

• Ignition for first-boot declarative config — consider cloud-init equivalent
• ChromeOS-style update engine is battle-tested
• Dynamic kernel module loading — Tiny Core's extension system provides similar flexibility

From Kairos:

• Container-based OS distribution (OCI images) — enables docker pull for OS updates
• P2P mesh clustering via libp2p — interesting for edge fleet bootstrapping
• Meta-distribution approach: don't reinvent, augment
• Static kernel+initrd shipped in container image = truly atomic full-stack updates

---

4. Architecture Design

4.1 High-Level Architecture

┌──────────────────────────────────────────────────────┐
│                    BOOT MEDIA                         │
│  ┌──────────┐  ┌──────────┐  ┌────────────────────┐ │
│  │ GRUB/    │  │ Partition│  │    Partition B      │ │
│  │ Syslinux │  │    A     │  │    (passive)        │ │
│  │ (EFI/    │  │ (active) │  │                     │ │
│  │  BIOS)   │  │          │  │  vmlinuz            │ │
│  │          │  │ vmlinuz  │  │  kubesolo-os.gz     │ │
│  │ Fallback │  │ kubesolo-│  │  extensions.tcz     │ │
│  │ counter  │  │ os.gz    │  │                     │ │
│  │          │  │ ext.tcz  │  │                     │ │
│  └──────────┘  └──────────┘  └────────────────────┘ │
│                                                      │
│  ┌──────────────────────────────────────────────────┐│
│  │              Persistent Data Partition            ││
│  │  /var/lib/kubesolo/  (K8s state, SQLite DB)      ││
│  │  /var/lib/containerd/ (container images/layers)  ││
│  │  /etc/kubesolo/       (node config)              ││
│  │  /var/log/            (logs, optional)           ││
│  │  /usr/local/          (user data)                ││
│  └──────────────────────────────────────────────────┘│
└──────────────────────────────────────────────────────┘

                    BOOT FLOW
                       │
          ┌────────────▼────────────┐
          │  GRUB loads vmlinuz +   │
          │  kubesolo-os.gz from    │
          │  active partition       │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  Kernel boots, mounts   │
          │  SquashFS root (ro)     │
          │  in RAM                 │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  init: mount persistent │
          │  partition, bind-mount  │
          │  writable paths         │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  Load kernel modules:   │
          │  br_netfilter, overlay, │
          │  ip_tables, veth        │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  Configure networking   │
          │  (cloud-init or static) │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  Start KubeSolo         │
          │  (single binary)        │
          └────────────┬────────────┘
                       │
          ┌────────────▼────────────┐
          │  K8s API available      │
          │  Node ready for         │
          │  workloads              │
          └─────────────────────────┘

4.2 Partition Layout

Disk Layout (minimum 8 GB recommended):

┌──────────────────────────────────────────────────────────┐
│ Partition 1: EFI/Boot (256 MB, FAT32)                    │
│   /EFI/BOOT/bootx64.efi (or /boot/grub for BIOS)        │
│   grub.cfg with A/B logic + fallback counter             │
├──────────────────────────────────────────────────────────┤
│ Partition 2: System A (512 MB, SquashFS image, read-only)│
│   vmlinuz                                                │
│   kubesolo-os.gz (initramfs: core.gz + KubeSolo ext)     │
├──────────────────────────────────────────────────────────┤
│ Partition 3: System B (512 MB, SquashFS image, read-only)│
│   (passive — receives updates, swaps with A)             │
├──────────────────────────────────────────────────────────┤
│ Partition 4: Persistent Data (remaining space, ext4)     │
│   /var/lib/kubesolo/     → K8s state, certs, SQLite      │
│   /var/lib/containerd/   → container images & layers     │
│   /etc/kubesolo/         → node configuration            │
│   /etc/network/          → network config                │
│   /var/log/              → system + K8s logs             │
│   /usr/local/            → user extensions               │
└──────────────────────────────────────────────────────────┘

4.3 Filesystem Mount Strategy

At boot, the init system constructs the runtime filesystem:

# Root: SquashFS from initramfs (read-only, in RAM)
/                          → tmpfs (RAM) + SquashFS overlay (ro)

# Persistent bind mounts from data partition
/var/lib/kubesolo          → /mnt/data/kubesolo        (rw)
/var/lib/containerd        → /mnt/data/containerd      (rw)
/etc/kubesolo              → /mnt/data/etc-kubesolo    (rw)
/etc/resolv.conf           → /mnt/data/resolv.conf     (rw)
/var/log                   → /mnt/data/log             (rw)
/usr/local                 → /mnt/data/usr-local       (rw)

# Everything else: read-only or tmpfs
/tmp                       → tmpfs
/run                       → tmpfs

4.4 Custom Initramfs (kubesolo-os.gz)

The initramfs is the core of the distribution — a remastered Tiny Core core.gz with KubeSolo baked in:

kubesolo-os.gz (cpio+gzip archive)
├── bin/                      # BusyBox symlinks
├── sbin/
│   └── init                  # Custom init script (see §4.5)
├── lib/
│   └── modules/              # Kernel modules (br_netfilter, overlay, etc.)
├── usr/
│   └── local/
│       └── bin/
│           └── kubesolo      # KubeSolo binary
├── opt/
│   ├── containerd/           # containerd + runc + CNI plugins
│   │   ├── bin/
│   │   │   ├── containerd
│   │   │   ├── containerd-shim-runc-v2
│   │   │   └── runc
│   │   └── cni/
│   │       └── bin/          # CNI plugins (bridge, host-local, loopback, etc.)
│   └── kubesolo-os/
│       ├── cloud-init.yaml   # Default cloud-init config
│       └── update-agent      # Atomic update agent binary
├── etc/
│   ├── os-release            # KubeSolo OS identification
│   ├── kubesolo/
│   │   └── config.yaml       # Default KubeSolo config
│   └── sysctl.d/
│       └── k8s.conf          # Kernel parameters for K8s
└── var/
    └── lib/
        └── kubesolo/         # Mount point (bind-mounted to persistent)

4.5 Init System

A custom init script replaces Tiny Core's default init to implement the appliance boot flow:

#!/bin/sh
# /sbin/init — KubeSolo OS init
set -e

# 1. Mount essential filesystems
mount -t proc proc /proc
mount -t sysfs sysfs /sys
mount -t devtmpfs devtmpfs /dev
mount -t tmpfs tmpfs /tmp
mount -t tmpfs tmpfs /run
mkdir -p /dev/pts /dev/shm
mount -t devpts devpts /dev/pts
mount -t tmpfs tmpfs /dev/shm

# 2. Parse boot parameters
PERSISTENT_DEV=""
for arg in $(cat /proc/cmdline); do
  case "$arg" in
    kubesolo.data=*)  PERSISTENT_DEV="${arg#kubesolo.data=}" ;;
    kubesolo.debug)   set -x ;;
    kubesolo.shell)   exec /bin/sh ;;  # Emergency shell
  esac
done

# 3. Mount persistent data partition
if [ -n "$PERSISTENT_DEV" ]; then
  mkdir -p /mnt/data
  # Wait for device (USB, slow disks)
  for i in $(seq 1 30); do
    [ -b "$PERSISTENT_DEV" ] && break
    sleep 1
  done
  mount -t ext4 "$PERSISTENT_DEV" /mnt/data

  # Create directory structure on first boot
  for dir in kubesolo containerd etc-kubesolo log usr-local network; do
    mkdir -p /mnt/data/$dir
  done

  # Bind mount persistent paths
  mount --bind /mnt/data/kubesolo     /var/lib/kubesolo
  mount --bind /mnt/data/containerd   /var/lib/containerd
  mount --bind /mnt/data/etc-kubesolo /etc/kubesolo
  mount --bind /mnt/data/log          /var/log
  mount --bind /mnt/data/usr-local    /usr/local
fi

# 4. Load required kernel modules
modprobe br_netfilter
modprobe overlay
modprobe ip_tables
modprobe iptable_nat
modprobe iptable_filter
modprobe veth
modprobe vxlan

# 5. Set kernel parameters
sysctl -w net.bridge.bridge-nf-call-iptables=1
sysctl -w net.bridge.bridge-nf-call-ip6tables=1
sysctl -w net.ipv4.ip_forward=1
sysctl -w fs.inotify.max_user_instances=1024
sysctl -w fs.inotify.max_user_watches=524288

# 6. Configure networking
#    Priority: cloud-init > persistent config > DHCP fallback
if [ -f /mnt/data/network/interfaces ]; then
  # Apply saved network config
  configure_network /mnt/data/network/interfaces
elif [ -f /mnt/data/etc-kubesolo/cloud-init.yaml ]; then
  # First boot: apply cloud-init
  apply_cloud_init /mnt/data/etc-kubesolo/cloud-init.yaml
else
  # Fallback: DHCP on first interface
  udhcpc -i eth0 -s /usr/share/udhcpc/default.script
fi

# 7. Set hostname
if [ -f /mnt/data/etc-kubesolo/hostname ]; then
  hostname $(cat /mnt/data/etc-kubesolo/hostname)
else
  hostname kubesolo-$(cat /sys/class/net/eth0/address | tr -d ':' | tail -c 6)
fi

# 8. Start containerd
containerd --config /etc/kubesolo/containerd-config.toml &
sleep 2  # Wait for socket

# 9. Start KubeSolo
exec /usr/local/bin/kubesolo \
  --path /var/lib/kubesolo \
  --local-storage true \
  $(cat /etc/kubesolo/extra-flags 2>/dev/null || true)

4.6 Atomic Update System

Update Flow

                    UPDATE PROCESS
                         │
           ┌─────────────▼──────────────┐
           │  1. Download new OS image   │
           │  (kubesolo-os-v2.img)       │
           │  Verify checksum + sig      │
           └─────────────┬──────────────┘
                         │
           ┌─────────────▼──────────────┐
           │  2. Write image to PASSIVE  │
           │  partition (B if A active)  │
           └─────────────┬──────────────┘
                         │
           ┌─────────────▼──────────────┐
           │  3. Update GRUB:           │
           │  - Set next boot → B       │
           │  - Set boot_counter = 3    │
           └─────────────┬──────────────┘
                         │
           ┌─────────────▼──────────────┐
           │  4. Reboot                 │
           └─────────────┬──────────────┘
                         │
           ┌─────────────▼──────────────┐
           │  5. GRUB boots partition B │
           │  Decrements boot_counter   │
           └─────────────┬──────────────┘
                         │
              ┌──────────┴──────────┐
              │                     │
        ┌─────▼─────┐        ┌─────▼─────┐
        │ Boot OK   │        │ Boot FAIL │
        │           │        │           │
        │ Health    │        │ Counter   │
        │ check OK  │        │ hits 0    │
        │           │        │           │
        │ Mark B as │        │ GRUB auto │
        │ default   │        │ rollback  │
        │ Clear     │        │ to A      │
        │ counter   │        │           │
        └───────────┘        └───────────┘

GRUB Configuration for A/B Boot

# /boot/grub/grub.cfg

set default=0
set timeout=3

# Saved environment variables:
#   active_slot    = A or B
#   boot_counter   = 3 (decremented each boot, 0 = rollback)
#   boot_success   = 0 (set to 1 by health check)

load_env

# If last boot failed and counter expired, swap slots
if [ "${boot_success}" != "1" ]; then
  if [ "${boot_counter}" = "0" ]; then
    if [ "${active_slot}" = "A" ]; then
      set active_slot=B
    else
      set active_slot=A
    fi
    save_env active_slot
    set boot_counter=3
    save_env boot_counter
  else
    # Decrement counter
    if [ "${boot_counter}" = "3" ]; then set boot_counter=2; fi
    if [ "${boot_counter}" = "2" ]; then set boot_counter=1; fi
    if [ "${boot_counter}" = "1" ]; then set boot_counter=0; fi
    save_env boot_counter
  fi
fi

set boot_success=0
save_env boot_success

# Boot from active slot
if [ "${active_slot}" = "A" ]; then
  set root=(hd0,gpt2)
else
  set root=(hd0,gpt3)
fi

menuentry "KubeSolo OS" {
  linux /vmlinuz kubesolo.data=/dev/sda4 quiet
  initrd /kubesolo-os.gz
}

menuentry "KubeSolo OS (emergency shell)" {
  linux /vmlinuz kubesolo.data=/dev/sda4 kubesolo.shell
  initrd /kubesolo-os.gz
}

Update Agent

A lightweight Go binary that runs as a Kubernetes CronJob or DaemonSet:

kubesolo-update-agent responsibilities:
  1. Poll update server (HTTPS) or watch OCI registry for new tags
  2. Download + verify new system image (SHA256 + optional GPG signature)
  3. Write to passive partition (dd or equivalent)
  4. Update GRUB environment (grub-editenv)
  5. Trigger reboot (via Kubernetes node drain → reboot)
  6. Post-boot health check:
     - KubeSolo API reachable?
     - containerd healthy?
     - Node Ready in kubectl?
     If all pass → set boot_success=1
     If any fail → leave boot_success=0 (auto-rollback on next reboot)

Update distribution models:

1. HTTP/S server — host images on a simple file server; agent polls for latest.json
2. OCI registry — tag system images as container images; agent pulls new tags
3. USB drive — for air-gapped: plug USB with new image, agent detects and applies
4. Portainer Edge — leverage existing Portainer Edge infrastructure for fleet updates

4.7 Configuration System

First Boot (cloud-init)

The system uses a simplified cloud-init compatible with Tiny Core's environment:

# /etc/kubesolo/cloud-init.yaml (placed on data partition before first boot)
#cloud-config

hostname: edge-node-001

network:
  version: 2
  ethernets:
    eth0:
      dhcp4: false
      addresses:
        - 192.168.1.100/24
      gateway4: 192.168.1.1
      nameservers:
        addresses:
          - 8.8.8.8
          - 1.1.1.1

kubesolo:
  extra-sans:
    - edge-node-001.local
    - 192.168.1.100
  local-storage: true
  portainer:
    edge-id: "your-edge-id"
    edge-key: "your-edge-key"

ssh:
  enabled: false  # Set true to enable SSH extension
  authorized_keys:
    - "ssh-rsa AAAA..."

ntp:
  servers:
    - pool.ntp.org

Runtime Configuration

Post-boot configuration changes via Kubernetes API:

# Access from the node (kubeconfig is at known path)
export KUBECONFIG=/var/lib/kubesolo/pki/admin/admin.kubeconfig
kubectl get nodes
kubectl apply -f workload.yaml

# Remote access via Portainer Edge or direct API
# (if apiserver-extra-sans includes remote IP/DNS)

---

5. Build Process

5.1 Build Pipeline

BUILD PIPELINE

┌─────────────────────┐     ┌──────────────────────┐
│ 1. Fetch Tiny Core  │────▶│ 2. Extract core.gz   │
│    Micro Core ISO   │     │    (cpio -idmv)      │
└─────────────────────┘     └──────────┬───────────┘
                                       │
                            ┌──────────▼───────────┐
                            │ 3. Inject KubeSolo   │
                            │    binary + deps     │
                            │    (containerd, runc,│
                            │     CNI, modules)    │
                            └──────────┬───────────┘
                                       │
                            ┌──────────▼───────────┐
                            │ 4. Replace /sbin/init│
                            │    with custom init  │
                            └──────────┬───────────┘
                                       │
                            ┌──────────▼───────────┐
                            │ 5. Repack initramfs  │
                            │    (find . | cpio -o │
                            │     | gzip > ks-os.gz│
                            └──────────┬───────────┘
                                       │
                            ┌──────────▼───────────┐
                            │ 6. Verify kernel has │
                            │    required configs  │
                            │    (cgroup v2, ns,   │
                            │     netfilter, etc.) │
                            └──────────┬───────────┘
                                       │
              ┌────────────────────────┼────────────────────┐
              │                        │                    │
    ┌─────────▼─────────┐   ┌─────────▼─────────┐  ┌──────▼───────┐
    │ 7a. Create ISO    │   │ 7b. Create raw    │  │ 7c. Create   │
    │     (bootable     │   │     disk image    │  │     OCI      │
    │      media)       │   │     (dd to disk)  │  │     image    │
    └───────────────────┘   └───────────────────┘  └──────────────┘

5.2 Build Script (Skeleton)

#!/bin/bash
# build-kubesolo-os.sh
set -euo pipefail

VERSION="${1:?Usage: $0 <version>}"
WORK_DIR="$(mktemp -d)"
OUTPUT_DIR="./output"

# --- 1. Download components ---
echo "==> Downloading Tiny Core Micro Core..."
wget -q "http://www.tinycorelinux.net/17.x/x86_64/release/CorePure64-17.0.iso" \
  -O "$WORK_DIR/core.iso"

echo "==> Downloading KubeSolo..."
curl -sfL https://get.kubesolo.io -o "$WORK_DIR/install-kubesolo.sh"
# Or: download specific release binary from GitHub

# --- 2. Extract Tiny Core ---
mkdir -p "$WORK_DIR/iso" "$WORK_DIR/rootfs"
mount -o loop "$WORK_DIR/core.iso" "$WORK_DIR/iso"
cp "$WORK_DIR/iso/boot/vmlinuz64" "$WORK_DIR/vmlinuz"
cd "$WORK_DIR/rootfs"
zcat "$WORK_DIR/iso/boot/corepure64.gz" | cpio -idmv 2>/dev/null
umount "$WORK_DIR/iso"

# --- 3. Inject KubeSolo + dependencies ---
# KubeSolo binary
mkdir -p usr/local/bin
cp /path/to/kubesolo usr/local/bin/kubesolo
chmod +x usr/local/bin/kubesolo

# containerd + runc + CNI (extracted from KubeSolo bundle or downloaded separately)
mkdir -p opt/cni/bin
# ... copy containerd, runc, CNI plugins

# Required kernel modules (if not already in core.gz)
# ... may need to compile or extract from Tiny Core extensions

# --- 4. Custom init ---
cat > sbin/init << 'INIT'
#!/bin/sh
# ... (init script from §4.5)
INIT
chmod +x sbin/init

# --- 5. Sysctl + OS metadata ---
mkdir -p etc/sysctl.d
cat > etc/sysctl.d/k8s.conf << EOF
net.bridge.bridge-nf-call-iptables = 1
net.bridge.bridge-nf-call-ip6tables = 1
net.ipv4.ip_forward = 1
fs.inotify.max_user_instances = 1024
fs.inotify.max_user_watches = 524288
EOF

cat > etc/os-release << EOF
NAME="KubeSolo OS"
VERSION="$VERSION"
ID=kubesolo-os
VERSION_ID=$VERSION
PRETTY_NAME="KubeSolo OS $VERSION"
HOME_URL="https://github.com/portainer/kubesolo"
EOF

# --- 6. Repack initramfs ---
find . | cpio -o -H newc 2>/dev/null | gzip -9 > "$WORK_DIR/kubesolo-os.gz"

# --- 7. Create disk image ---
mkdir -p "$OUTPUT_DIR"
create_disk_image "$WORK_DIR/vmlinuz" "$WORK_DIR/kubesolo-os.gz" \
  "$OUTPUT_DIR/kubesolo-os-${VERSION}.img"

echo "==> Built: $OUTPUT_DIR/kubesolo-os-${VERSION}.img"

5.3 Alternative: Kairos-based Build (Container-first)

For faster iteration, leverage the Kairos framework to get A/B updates, P2P mesh, and OCI distribution for free:

# Dockerfile.kubesolo-os
FROM quay.io/kairos/core-alpine:latest

# Install KubeSolo
RUN curl -sfL https://get.kubesolo.io | sh -

# Pre-configure
COPY kubesolo-config.yaml /etc/kubesolo/config.yaml
COPY cloud-init-defaults.yaml /system/oem/

# Kernel modules
RUN apk add --no-cache \
  linux-lts \
  iptables \
  iproute2 \
  conntrack-tools

# Sysctl
COPY k8s-sysctl.conf /etc/sysctl.d/

# Build: docker build -t kubesolo-os:v1 .
# Flash: Use AuroraBoot or Kairos tooling to convert OCI → bootable image

Advantages of Kairos approach:

• A/B atomic updates with rollback — built-in
• OCI-based distribution — docker push your OS
• P2P mesh bootstrapping — nodes find each other
• Kubernetes-native upgrades — kubectl apply to upgrade the OS
• Proven in production edge deployments

Trade-offs:

• Larger footprint than pure Tiny Core remaster (~200–400 MB vs ~50–80 MB)
• Dependency on Kairos project maintenance
• Less control over boot process internals

---

6. Kernel Considerations

6.1 Tiny Core Kernel Audit

Tiny Core 17.0 ships a modern 6.x kernel, which should have cgroup v2 support compiled in. However, the kernel config must be verified for these critical options:

# MANDATORY for KubeSolo
CONFIG_CGROUPS=y
CONFIG_CGROUP_CPUACCT=y
CONFIG_CGROUP_DEVICE=y
CONFIG_CGROUP_FREEZER=y
CONFIG_CGROUP_SCHED=y
CONFIG_CGROUP_PIDS=y
CONFIG_MEMCG=y
CONFIG_CGROUP_BPF=y

CONFIG_NAMESPACES=y
CONFIG_NET_NS=y
CONFIG_PID_NS=y
CONFIG_USER_NS=y
CONFIG_UTS_NS=y
CONFIG_IPC_NS=y

CONFIG_OVERLAY_FS=y      # or =m (module)
CONFIG_BRIDGE=y           # or =m
CONFIG_NETFILTER=y
CONFIG_NF_NAT=y
CONFIG_IP_NF_IPTABLES=y
CONFIG_IP_NF_NAT=y
CONFIG_IP_NF_FILTER=y
CONFIG_VETH=y             # or =m
CONFIG_VXLAN=y            # or =m

CONFIG_SQUASHFS=y         # For Tiny Core's own extension system
CONFIG_BLK_DEV_LOOP=y     # For SquashFS mounting

# RECOMMENDED
CONFIG_BPF_SYSCALL=y      # For modern CNI plugins
CONFIG_CRYPTO_SHA256=y    # For image verification
CONFIG_SECCOMP=y          # Container security
CONFIG_AUDIT=y            # Audit logging

If the stock Tiny Core kernel lacks any of these, options are:

1. Load as modules — if compiled as =m, load via modprobe in init
2. Recompile kernel — use Tiny Core's kernel build process with custom config
3. Use a different kernel — e.g., pull the kernel from Alpine Linux or build from mainline

6.2 Custom Kernel Build (if needed)

# On a Tiny Core build system
tce-load -wi compiletc linux-6.x-source
cd /usr/src/linux-6.x
cp /path/to/kubesolo-kernel.config .config
make oldconfig
make -j$(nproc) bzImage modules
# Extract vmlinuz and required modules

---

7. Security Model

7.1 Layered Security

┌─────────────────────────────────────────┐
│        APPLICATION LAYER                │
│  Kubernetes RBAC + Network Policies     │
│  Pod Security Standards                 │
│  Seccomp / AppArmor profiles            │
├─────────────────────────────────────────┤
│        CONTAINER RUNTIME LAYER          │
│  containerd with default seccomp        │
│  Read-only container rootfs             │
│  User namespace mapping (optional)      │
├─────────────────────────────────────────┤
│        OS LAYER                         │
│  SquashFS root (read-only, in RAM)      │
│  No package manager                     │
│  No SSH by default                      │
│  Minimal userland (BusyBox only)        │
│  No compiler, no debugger               │
├─────────────────────────────────────────┤
│        BOOT LAYER                       │
│  Signed images (GPG verification)       │
│  Secure Boot (optional, UEFI)           │
│  A/B rollback on tamper/failure         │
│  TPM-based attestation (optional)       │
└─────────────────────────────────────────┘

7.2 Attack Surface Comparison

Attack Vector	Traditional Linux	KubeSolo OS
Package manager exploit	Possible (apt/yum)	Eliminated (no pkg manager)
SSH brute force	Common	Eliminated (no SSH default)
Writable system files	Yes (/etc, /usr)	Eliminated (SquashFS ro)
Persistent rootkit	Survives reboot	Eliminated (RAM-only root)
Kernel module injection	Possible	Mitigated (only preloaded modules)
Local privilege escalation	Various paths	Reduced (minimal binaries)

---

8. Implementation Roadmap

Phase 1 — Proof of Concept (2–3 weeks)

Goal: Boot Tiny Core + KubeSolo, validate K8s functionality.

1. Download Tiny Core Micro Core 17.0 (x86_64)
2. Extract core.gz, inject KubeSolo binary
3. Create custom init that starts KubeSolo
4. Verify kernel has required configs (cgroup v2, namespaces, netfilter)
5. Build bootable ISO, test in QEMU/KVM
6. Deploy a test workload (nginx pod)
7. Validate: kubectl get nodes shows Ready

Success criteria: Single ISO boots to functional K8s node in < 30 seconds.

Phase 2 — Persistence + Immutability (2–3 weeks)

Goal: Persistent K8s state across reboots, immutable root.

1. Implement persistent data partition with bind mounts
2. Verify K8s state survives reboot (pods, services, PVCs)
3. Verify SQLite DB integrity across unclean shutdowns
4. Lock down root filesystem (verify read-only enforcement)
5. Test: corrupt system files → verify RAM-only root is unaffected

Phase 3 — Atomic Updates + Rollback (3–4 weeks)

Goal: A/B partition updates with automatic rollback.

1. Implement GRUB A/B boot configuration
2. Build update agent (Go binary)
3. Implement health check + boot_success flag
4. Test update cycle: A → B → verify → mark good
5. Test rollback: A → B → fail → auto-revert to A
6. Test: pull power during update → verify clean state

Phase 4 — Production Hardening (2–3 weeks)

Goal: Production-ready security and manageability.

1. Image signing and verification (GPG or sigstore/cosign)
2. Cloud-init implementation for first-boot config
3. Portainer Edge integration testing
4. Optional SSH extension (.tcz)
5. Optional management API (lightweight, mTLS-authenticated)
6. Performance benchmarking (boot time, memory usage, disk I/O)
7. Documentation and deployment guides

Phase 5 — Distribution + Fleet Management (ongoing)

Goal: Scale to fleet deployments.

1. CI/CD pipeline for automated image builds
2. OCI registry distribution (optional)
3. Fleet update orchestration (rolling updates across nodes)
4. Monitoring integration (Prometheus metrics endpoint)
5. USB provisioning tool for air-gapped deployments
6. ARM64 support (Raspberry Pi, Jetson, etc.)

---

9. Open Questions & Decisions

#	Question	Options	Recommendation
1	Build approach	Pure Tiny Core remaster vs. Kairos framework	Start with pure remaster for minimal footprint; evaluate Kairos if update complexity becomes unmanageable
2	Kernel	Stock Tiny Core kernel vs. custom build	Audit stock kernel first; only custom-build if missing critical configs
3	Management interface	SSH / API / Portainer Edge only	Portainer Edge primary; optional SSH extension for debugging
4	Update distribution	HTTP server / OCI registry / USB	HTTP for simplicity; OCI if leveraging container infrastructure
5	Init system	Custom shell script vs. BusyBox init vs. s6	Custom shell script for PoC; evaluate s6 for supervision
6	Networking	DHCP only / Static / cloud-init	Cloud-init with DHCP fallback
7	Architecture support	x86_64 only vs. multi-arch	x86_64 first; ARM64 in Phase 5
8	Container images	Preloaded in initramfs vs. pull at boot	Preload core workloads; pull additional at runtime

---

10. References

• KubeSolo: https://github.com/portainer/kubesolo
• Tiny Core Linux: http://www.tinycorelinux.net
• Tiny Core Wiki (Remastering): http://wiki.tinycorelinux.net/doku.php?id=wiki:remastering
• Talos Linux: https://www.talos.dev
• Kairos: https://kairos.io
• Bottlerocket: https://github.com/bottlerocket-os/bottlerocket
• Flatcar Linux: https://www.flatcar.org
• Kubernetes Node Requirements: https://kubernetes.io/docs/setup/production-environment/container-runtimes/
• cgroup v2: https://www.kernel.org/doc/html/latest/admin-guide/cgroup-v2.html