FreeBSD NAS Building Guide: Hardware to Software

Building a NAS (Network Attached Storage) on FreeBSD is one of the most rewarding infrastructure projects you can undertake. FreeBSD's native ZFS support, stable networking stack, and lightweight base system make it an ideal NAS operating system. Unlike appliance-based solutions, a FreeBSD NAS gives you full control over hardware selection, filesystem configuration, and service deployment.

This guide walks through every step: selecting hardware (ECC RAM, HBA cards, cases, drives), installing FreeBSD, configuring ZFS for NAS workloads, setting up file sharing with Samba and NFS, monitoring the system, and integrating a UPS for power protection.

Hardware Selection

CPU

NAS workloads are primarily I/O-bound, not CPU-bound. A modern CPU spends most of its time waiting for disks. However, ZFS compression and checksumming do use CPU cycles, and if you plan to run services in jails (Plex, Nextcloud, etc.), more CPU helps.

Budget NAS (home/small office): Intel Core i3 or AMD Ryzen 3. Four cores are sufficient.

Mid-range NAS (small business, 10+ users): Intel Core i5 or AMD Ryzen 5. Handles Samba, NFS, ZFS scrubs, and jails.

Enterprise/prosumer: Intel Xeon E or AMD EPYC for ECC memory support and more PCIe lanes.

ECC RAM

Use ECC RAM. This is the strongest hardware recommendation in this guide.

ECC (Error-Correcting Code) RAM detects and corrects single-bit memory errors. ZFS checksums protect data on disk, but if a bit flips in RAM between reading from disk and writing to the network (or vice versa), ZFS cannot detect it. ECC RAM closes this gap.

Minimum: 8 GB. Recommended: 16-32 GB. Rule of thumb: 1 GB of ARC per TB of raw storage. Consumer platforms (Intel Core, AMD Ryzen on non-Pro chipsets) generally do not support ECC -- use Xeon E, Ryzen Pro, or EPYC. If ECC is not feasible, ZFS on non-ECC is still safer than any other filesystem on non-ECC. But ECC is the correct choice for data you care about.

HBA (Host Bus Adapter)

The storage controller is critical. Do not use hardware RAID controllers. ZFS needs direct access to the raw disks. Hardware RAID controllers hide disk errors from ZFS, defeating its self-healing capabilities.

Use an HBA (IT mode) card that presents each disk individually to the OS.

Recommended HBAs:

LSI SAS 9211-8i (rebranded as Broadcom): The gold standard for FreeBSD NAS builds. Supports 8 SAS/SATA drives. Available for $20-40 used. Flash to IT mode firmware if it comes in IR (RAID) mode.
LSI SAS 9300-8i: Newer generation, 12 Gbps SAS. Same reliability, higher throughput for SAS SSDs.
LSI SAS 9207-8i: Similar to 9211, good FreeBSD compatibility.

Flashing to IT mode (if needed):

sh
# This is done before FreeBSD installation, typically from an EFI shell
# Download IT mode firmware from Broadcom/Avago
# Flash using sas2flash utility
sas2flash -o -f 2118it.bin -b mptsas2.rom

Drives

For bulk storage (data pool):

WD Red Plus (CMR) or WD Ultrastar for NAS duty
Seagate IronWolf or Exos for enterprise workloads
Avoid SMR (Shingled Magnetic Recording) drives -- they have poor random write performance that degrades ZFS resilver times

For cache (L2ARC/SLOG):

Intel Optane (if available used) for SLOG (ZFS Intent Log)
Enterprise NVMe SSDs (Intel DC, Samsung PM series) for L2ARC
Consumer NVMe works but check write endurance ratings

For boot:

A pair of mirrored small SSDs (120-240 GB) or USB drives
Keep the boot pool separate from the data pool

Case

NAS cases need drive bays, airflow, and enough space for the motherboard and HBA.

Budget: Fractal Design Node 304 (6 x 3.5" bays, Mini-ITX)

Mid-range: Fractal Design Define 7 (up to 14 x 3.5" bays, ATX)

Rackmount: Supermicro 846 chassis (24 x 3.5" bays, hot-swap) or Rosewill RSV-L4500U (15 x 3.5" bays)

Network

1 GbE: Sufficient for home use with 1-3 clients
2.5 GbE: Good price/performance for small offices. Intel I225-V or Realtek RTL8125
10 GbE: Required for video editing, VM storage, or serving many clients. Intel X550-T2 or Mellanox ConnectX-3/4

FreeBSD has excellent Intel NIC driver support. Intel is preferred over Realtek for stability.

FreeBSD Installation

Download the FreeBSD installer, write to USB with dd, boot, and select ZFS as the root filesystem with a mirrored boot pool on the boot SSDs.

Post-Installation Essentials

sh
# Update the base system
freebsd-update fetch install

# Install essential packages
pkg install smartmontools tmux htop

# Enable SSH
sysrc sshd_enable="YES"
service sshd start

# Set timezone
tzsetup

# Enable NTP
sysrc ntpd_enable="YES"
service ntpd start

Kernel Tuning for NAS

Edit /boot/loader.conf:

sh
# ZFS ARC maximum (adjust to your RAM, leave 4-8GB for OS and services)
vfs.zfs.arc_max="25769803776"  # 24 GB for a 32 GB system

# Increase AIO threads for Samba
vfs.aio.max_aio_queue_per_proc=4096
vfs.aio.max_aio_per_proc=2048

# Enable AHCI for SATA controllers
ahci_load="YES"

# Load HBA driver at boot
mps_load="YES"    # For LSI 9211
mpr_load="YES"    # For LSI 9300

Edit /etc/sysctl.conf:

sh
# Network tuning for file serving
net.inet.tcp.sendbuf_max=16777216
net.inet.tcp.recvbuf_max=16777216
kern.ipc.maxsockbuf=16777216
net.inet.tcp.sendbuf_auto=1
net.inet.tcp.recvbuf_auto=1

# Increase maximum open files
kern.maxfiles=65536
kern.maxfilesperproc=32768

ZFS Layout

Creating the Data Pool

For a NAS with 6 drives, RAID-Z2 provides a good balance of space and redundancy:

sh
# Create the data pool
zpool create -o ashift=12 \
    -O compression=lz4 \
    -O atime=off \
    -O xattr=sa \
    -O acltype=posixacl \
    datapool raidz2 da0 da1 da2 da3 da4 da5

Key options: ashift=12 matches modern 4K-sector drives, compression=lz4 provides near-zero CPU overhead with space savings, atime=off reduces write I/O, xattr=sa and acltype=posixacl enable Samba ACL support.

Dataset Structure

Create separate datasets for different purposes:

sh
# Shared folders
zfs create datapool/shared
zfs create datapool/shared/documents
zfs create datapool/shared/media
zfs create datapool/shared/photos
zfs create datapool/shared/backups

# Per-user home directories
zfs create datapool/home
zfs create datapool/home/alice
zfs create datapool/home/bob

# Set quotas
zfs set quota=500G datapool/shared/media
zfs set quota=100G datapool/home/alice
zfs set quota=100G datapool/home/bob

# Tune for media (large sequential files)
zfs set recordsize=1M datapool/shared/media
zfs set recordsize=1M datapool/shared/photos

# Tune for documents (smaller files)
zfs set recordsize=128K datapool/shared/documents

Adding Cache Devices

sh
# Add L2ARC (read cache) -- useful if ARC is full and read latency matters
zpool add datapool cache nvd0

# Add SLOG (write log) -- critical for synchronous writes (NFS, databases)
zpool add datapool log mirror nvd1 nvd2

Automated Snapshots

sh
pkg install zfstools

Or use a cron-based approach:

sh
# /etc/cron.d/zfs-snapshots
15 * * * *  root  zfs snapshot -r datapool@auto-hourly-$(date +\%Y\%m\%d-\%H\%M)
0 0 * * *   root  zfs snapshot -r datapool@auto-daily-$(date +\%Y\%m\%d)
0 0 * * 0   root  zfs snapshot -r datapool@auto-weekly-$(date +\%Y\%m\%d)

# Cleanup: keep 48 hourly, 30 daily, 12 weekly
30 0 * * *  root  zfs list -H -t snapshot -o name | grep auto-hourly | sort | head -n -48 | xargs -n1 zfs destroy 2>/dev/null
30 0 * * *  root  zfs list -H -t snapshot -o name | grep auto-daily | sort | head -n -30 | xargs -n1 zfs destroy 2>/dev/null
30 0 * * 0  root  zfs list -H -t snapshot -o name | grep auto-weekly | sort | head -n -12 | xargs -n1 zfs destroy 2>/dev/null

Scheduled Scrubs

sh
# Weekly scrub (add to /etc/cron.d/zfs-scrub)
0 2 * * 6   root  zpool scrub datapool

Samba provides Windows-compatible file sharing (SMB/CIFS). It is the primary sharing protocol for mixed networks with Windows, macOS, and Linux clients.

Installation

sh
pkg install samba419
sysrc samba_server_enable="YES"

Configuration

Create /usr/local/etc/smb4.conf:

sh
[global]
    workgroup = WORKGROUP
    server string = FreeBSD NAS
    server role = standalone server

    # Security
    security = user
    passdb backend = tdbsam
    map to guest = Bad User

    # Performance
    socket options = TCP_NODELAY IPTOS_LOWDELAY
    read raw = yes
    write raw = yes
    use sendfile = yes
    aio read size = 16384
    aio write size = 16384
    min receivefile size = 16384

    # Logging
    log file = /var/log/samba4/log.%m
    max log size = 1000
    log level = 1

    # macOS compatibility
    vfs objects = catia fruit streams_xattr
    fruit:metadata = stream
    fruit:model = MacSamba

[documents]
    path = /datapool/shared/documents
    valid users = @staff
    browseable = yes
    writable = yes
    create mask = 0664
    directory mask = 0775

[media]
    path = /datapool/shared/media
    valid users = @staff
    browseable = yes
    read only = yes
    write list = alice bob

[photos]
    path = /datapool/shared/photos
    valid users = @family
    browseable = yes
    writable = yes
    create mask = 0664
    directory mask = 0775

[homes]
    comment = Home Directories
    browseable = no
    writable = yes
    create mask = 0600
    directory mask = 0700

User Setup

sh
# Create system users
pw useradd alice -m -s /usr/sbin/nologin
pw useradd bob -m -s /usr/sbin/nologin

# Create groups
pw groupadd staff
pw groupmod staff -m alice,bob
pw groupadd family
pw groupmod family -m alice,bob

# Set Samba passwords
smbpasswd -a alice
smbpasswd -a bob

# Set filesystem ownership
chown -R alice:staff /datapool/shared/documents
chmod -R 2775 /datapool/shared/documents

Start Samba:

sh
service samba_server start

Test the configuration:

sh
testparm /usr/local/etc/smb4.conf

NFS is the standard Unix/Linux file sharing protocol. Use it for sharing with other FreeBSD, Linux, or macOS clients (macOS supports both SMB and NFS).

Configuration

Enable NFS server services:

sh
sysrc nfs_server_enable="YES"
sysrc nfsv4_server_enable="YES"
sysrc nfsuserd_enable="YES"
sysrc mountd_enable="YES"
sysrc rpcbind_enable="YES"

Configure exports in /etc/exports:

sh
# /etc/exports
# Share to specific subnet
/datapool/shared/documents -network 10.0.1.0 -mask 255.255.255.0
/datapool/shared/media -network 10.0.1.0 -mask 255.255.255.0 -ro

# Share to specific hosts
/datapool/home/alice -maproot=alice alice-workstation.local
/datapool/home/bob -maproot=bob bob-workstation.local

# NFSv4 root export
V4: /datapool -sec=sys -network 10.0.1.0 -mask 255.255.255.0

Start NFS services:

sh
service nfsd start
service mountd start
service rpcbind start
service nfsuserd start

Verify exports:

sh
showmount -e localhost

Monitoring

SMART Monitoring

Monitor drive health with smartmontools:

sh
pkg install smartmontools
sysrc smartd_enable="YES"

Configure /usr/local/etc/smartd.conf:

sh
# Monitor all drives, send email on issues
DEVICESCAN -a -o on -S on -n standby,q -s (S/../.././02|L/../../6/03) -W 4,45,55 -m admin@example.com

sh
service smartd start

ZFS Health Monitoring

Create a monitoring script:

sh
#!/bin/sh
# /usr/local/bin/nas-monitor.sh

# Check pool health
POOL_HEALTH=$(zpool status -x)
if [ "$POOL_HEALTH" != "all pools are healthy" ]; then
    echo "WARNING: ZFS pool issue detected"
    zpool status
    echo "$POOL_HEALTH" | mail -s "NAS ZFS Alert: Pool Degraded" admin@example.com
fi

# Check disk space
zpool list -H -o name,capacity | while IFS=$'\t' read name cap; do
    cap_num=$(echo "$cap" | tr -d '%')
    if [ "$cap_num" -gt 80 ]; then
        echo "WARNING: Pool $name is ${cap} full" | mail -s "NAS Alert: Pool $name space low" admin@example.com
    fi
done

# Check drive temperatures
for drive in /dev/da[0-9]*; do
    temp=$(smartctl -A "$drive" 2>/dev/null | awk '/Temperature_Celsius/{print $10}')
    if [ -n "$temp" ] && [ "$temp" -gt 50 ]; then
        echo "WARNING: $drive temperature is ${temp}C" | mail -s "NAS Alert: Drive temperature high" admin@example.com
    fi
done

Schedule it:

sh
chmod +x /usr/local/bin/nas-monitor.sh
# Run every 30 minutes
echo "*/30 * * * * root /usr/local/bin/nas-monitor.sh" >> /etc/crontab

Prometheus Integration

For Grafana dashboards, install node_exporter:

sh
pkg install node_exporter
sysrc node_exporter_enable="YES"
sysrc node_exporter_args="--collector.zfs --collector.textfile --collector.textfile.directory=/var/tmp/node_exporter"
service node_exporter start

UPS Integration

A UPS (Uninterruptible Power Supply) is not optional for a NAS. Power loss during a ZFS write can corrupt the ZIL or in-flight data. While ZFS is designed to survive power loss, a graceful shutdown is always preferred.

NUT (Network UPS Tools)

sh
pkg install nut

Configure for a USB-connected UPS:

sh
# /usr/local/etc/nut/ups.conf
[myups]
    driver = usbhid-ups
    port = auto
    desc = "Server Room UPS"

sh
# /usr/local/etc/nut/upsd.conf
LISTEN 127.0.0.1 3493

sh
# /usr/local/etc/nut/upsd.users
[admin]
    password = secret
    actions = set
    instcmds = all

[upsmon]
    password = secret
    upsmon master

sh
# /usr/local/etc/nut/upsmon.conf
MONITOR myups@localhost 1 upsmon secret master
SHUTDOWNCMD "/sbin/shutdown -p now"
POWERDOWNFLAG /etc/killpower
NOTIFYCMD "/usr/local/sbin/upssched"

sh
# /usr/local/etc/nut/nut.conf
MODE=standalone

Enable and start:

sh
sysrc nut_enable="YES"
sysrc nut_upsmon_enable="YES"
sysrc nut_upslog_enable="YES"
service nut start
service nut_upsmon start

Verify UPS communication:

sh
upsc myups@localhost

This shows battery level, load, input voltage, and runtime estimate.

NUT handles automatic shutdown when battery runs low. The defaults shut down when battery reaches critical level. Verify with upsc myups@localhost that battery and runtime metrics are being read correctly.

FAQ

Do I really need ECC RAM for a FreeBSD NAS?

For data you care about, yes. ZFS provides end-to-end data integrity on disk, but a bit flip in RAM between read and write bypasses all ZFS protections. ECC RAM corrects single-bit errors and detects double-bit errors. The price premium for ECC is small compared to the cost of silent data corruption. If budget forces a choice between ECC and more storage, choose ECC.

How many drives should I start with?

For a first NAS, 4 drives in RAID-Z1 or a pair of mirrors gives you a good starting point. RAID-Z2 requires a minimum of 4 drives but 6 is more practical (4 for data, 2 for parity). You can expand later by adding vdevs (mirror pairs or additional RAID-Z groups), but you cannot add individual disks to an existing RAID-Z vdev (except with the new RAID-Z expansion feature in OpenZFS 2.2+).

Should I use a dedicated boot drive?

Yes. Keep the boot pool on separate drives (a pair of mirrored SSDs or USB sticks) from the data pool. This allows you to reinstall or upgrade FreeBSD without touching your data. If a boot drive fails, you still have the data pool intact.

What is the performance difference between RAID-Z1, RAID-Z2, and mirrors?

Mirrors provide the best random read performance (reads can be served from any mirror member) and the fastest resilver times. RAID-Z2 provides better space efficiency but lower random read IOPS. For a NAS serving files to a small number of clients, the difference is negligible. For a NAS serving VMs or databases, mirrors are preferred.

How do I replace a failed drive?

sh
# Identify the failed drive
zpool status datapool

# Replace it (da1 is failed, da6 is the new drive)
zpool replace datapool da1 da6

# Monitor resilver progress
zpool status datapool

The pool remains online during resilver. Performance is reduced but all data remains accessible.

Enable the ZFS .zfs directory:

sh
zfs set snapdir=visible datapool/shared/documents

Users can then browse \\NAS\documents\.zfs\snapshot\ via Samba to find and recover their own files from any snapshot. This is the FreeBSD equivalent of Windows "Previous Versions."