Backup & Disaster Recovery — because snapshots are not backups.

Say it again: SNAPSHOTS ARE NOT BACKUPS. Snapshots protect against accidental deletion and logical corruption. They do NOT protect against hardware failure, pool corruption, or site-wide disasters. If the pool dies, all snapshots die with it. You need zfs send/recv to a separate system. Here's how.

The recipe

# On the backup server (another kldload box, TrueNAS, or any ZFS system)
kdir /srv/backups

# Create per-host datasets
kdir /srv/backups/web-prod-01
kdir /srv/backups/db-prod-01

# On the production server
kpkg install sanoid

# Configure snapshot policy
cat > /etc/sanoid/sanoid.conf <<'SANOID'
[rpool/ROOT/kldload-node]
  use_template = production
  recursive = yes

[rpool/home]
  use_template = production
  recursive = yes

[rpool/srv]
  use_template = production
  recursive = yes

[template_production]
  autosnap = yes
  autoprune = yes
  hourly = 24
  daily = 30
  weekly = 8
  monthly = 12
  yearly = 2
SANOID

# Enable the timer
systemctl enable --now sanoid.timer

# One-time initial sync (full send)
syncoid --recursive rpool/home backup-server:tank/backups/web-prod-01/home

# Cron job for ongoing incremental replication
# Every hour, only changed blocks are sent
cat > /etc/cron.d/syncoid-backup <<'CRON'
0 * * * * root syncoid --recursive --no-sync-snap rpool/home backup-server:tank/backups/$(hostname)/home >> /var/log/syncoid.log 2>&1
15 * * * * root syncoid --recursive --no-sync-snap rpool/srv backup-server:tank/backups/$(hostname)/srv >> /var/log/syncoid.log 2>&1
30 * * * * root syncoid --recursive --no-sync-snap rpool/ROOT backup-server:tank/backups/$(hostname)/ROOT >> /var/log/syncoid.log 2>&1
CRON

# Replicate encrypted datasets without decrypting
# The backup server stores ciphertext — can't read your data
syncoid --recursive --sendoptions="w" \
    rpool/home offsite-server:tank/offsite/$(hostname)/home

# Or send to cloud storage (S3-compatible)
zfs send -w rpool/home@daily-$(date +%Y%m%d) | \
    aws s3 cp - s3://my-bucket/backups/$(hostname)/home-$(date +%Y%m%d).zfs

# Even AWS can't read your data — keys never leave your machine

# SCENARIO: Production server is dead. Recover from backup.

# 1. Boot a new kldload system (USB or VM)

# 2. Create the pool
zpool create -o ashift=12 rpool /dev/vda2

# 3. Receive the backup
ssh backup-server "zfs send -R tank/backups/web-prod-01/ROOT@latest" | \
    zfs recv -F rpool/ROOT
ssh backup-server "zfs send -R tank/backups/web-prod-01/home@latest" | \
    zfs recv -F rpool/home
ssh backup-server "zfs send -R tank/backups/web-prod-01/srv@latest" | \
    zfs recv -F rpool/srv

# 4. Install bootloader
krecovery reinstall-bootloader /dev/vda

# 5. Reboot — you're back online
# Data loss = time since last syncoid run (worst case: 1 hour)

The 3-2-1 rule with ZFS