Docker Data Root Migration to External Disk

Problem

Elasticsearch listings_general index hit the flood-stage disk watermark (95%+) on root partition /dev/sda2 (32GB, 94% used). ES automatically set the index to read-only-allow-delete, blocking all writes (indexing new listings failed with cluster_block_exception).

Root cause: Docker data (/var/lib/docker) lived on the small root partition.

Options Considered

Choice: Option C

Reason: NVMe SSD (/dev/nvme0n1p2, 476GB) contained Windows files that could not be lost. Option C allows using the disk for Docker without touching existing data.

Steps Executed

1. Install NTFS support

sudo pacman -S ntfs-3g

• ntfs-3g: userspace NTFS driver for Linux, enables read/write access to NTFS partitions.

2. Mount the NVMe NTFS partition

sudo mkdir -p /mnt/nvme
sudo mount -t ntfs-3g /dev/nvme0n1p2 /mnt/nvme

• mkdir -p: create directory and any missing parents; no error if it already exists.
• mount -t ntfs-3g: mount using the NTFS-3G driver.

3. Create a 100GB image file

sudo dd if=/dev/zero of=/mnt/nvme/docker.img bs=1M count=102400 status=progress

• dd: low-level byte copying tool.
• if=/dev/zero: input is an infinite stream of zero bytes.
• of=...: output file path.
• bs=1M: block size of 1 megabyte.
• count=102400: write 102400 blocks = 100GB.
• status=progress: show transfer progress during copy.

4. Format the image as ext4

sudo mkfs.ext4 /mnt/nvme/docker.img

• mkfs.ext4: create an ext4 filesystem inside the image file. This does NOT affect the NTFS partition — it only formats the contents of the image file.

5. Loop-mount the image

sudo mkdir -p /mnt/docker
sudo mount -o loop /mnt/nvme/docker.img /mnt/docker

• mount -o loop: treat a regular file as a block device (loop device) and mount it. The kernel creates /dev/loopN automatically.

6. Stop Docker and copy data

docker stop $(docker ps -q)
sudo systemctl stop docker
sudo systemctl stop docker.socket
sudo rsync -aP /var/lib/docker/ /mnt/docker/

• docker ps -q: list running container IDs (quiet mode).
• docker stop $(...): stop all running containers before stopping the daemon.
• systemctl stop docker.socket: stop the Docker socket (prevents auto-restart of daemon).
• rsync -aP: archive mode (-a preserves permissions, ownership, timestamps, symlinks) + progress (-P).

7. Configure Docker to use new data root

sudo tee /etc/docker/daemon.json <<'EOF'
{
  "data-root": "/mnt/docker"
}
EOF

• daemon.json: Docker daemon configuration file.
• data-root: tells Docker where to store images, containers, volumes, etc.

8. Start Docker and verify

sudo systemctl start docker
docker info | grep "Docker Root Dir"
# Output: Docker Root Dir: /mnt/docker
df -h /mnt/docker
# Output: /dev/loop0  98G  1.3G  92G  2%  /mnt/docker

9. Persist mounts across reboots

sudo blkid /dev/nvme0n1p2
# UUID=26CCDAA5CCDA6F15

echo "UUID=26CCDAA5CCDA6F15 /mnt/nvme ntfs-3g defaults,nofail 0 0" | sudo tee -a /etc/fstab
echo "/mnt/nvme/docker.img /mnt/docker ext4 loop,defaults,nofail 0 0" | sudo tee -a /etc/fstab

• /etc/fstab: filesystem table; defines mounts that are applied at boot.
• nofail: system continues booting even if this mount fails (e.g., disk not plugged in).
• 0 0: no dump backup, no fsck ordering.

10. Fix Elasticsearch index block

curl -X PUT "localhost:9200/listings_general/_settings" \
  -H 'Content-Type: application/json' \
  -d '{"index.blocks.read_only_allow_delete": null}'

• Removes the read-only block that ES applied when disk was full.

11. (Optional) Clean up old Docker data

sudo rm -rf /var/lib/docker

• Frees space on root partition after confirming Docker works from the new location.

Key Concepts

Disk Layout After Migration

/dev/sda2 (32G, root /)        -> OS + apps (no longer holds Docker data)
/dev/nvme0n1p2 (476G, NTFS)    -> mounted at /mnt/nvme (Windows files preserved)
  └── docker.img (100G, ext4)  -> loop-mounted at /mnt/docker (Docker data-root)

written on: 4 . 14 . 2026 xyz