Files
linuxkit/docs/external-disk.md
Rolf Neugebauer f90fde5a6f cmd: Unify disk command line options for local hypervisors
- '-disk-size' is now defaults to MB (but can be GB when appending 'G')
- The disk will be created if it doesn't exist (didn't happen in qemu)

Update the documentation.

Signed-off-by: Rolf Neugebauer <rolf.neugebauer@docker.com>
2017-05-24 16:05:06 +01:00

3.0 KiB

External Disk

linuxkit run has the ability to mount an external disk when booting. It involves two steps:

  1. Make the disk available as a device
  2. Mount the disk

Make Disk Available

In order to make the disk available, you need to tell linuxkit where the disk file or block device is.

All local linuxkit run methods (currently hyperkit, qemu, and vmware) take two arguments:

  • -disk-size <size>: size of disk. The default is in MB but a G can be appended to specify the size in GB, e.g. -disk-size 4096 or disk-size 4G will provide a 4GB disk.
  • -disk <path>: use the disk at location path, e.g. -disk foo.img will use the disk at $PWD/foo.img

If you do not provide -disk path, linuxkit assumes a default, which is prefix-state/disk.img for hyperkit and vmware and prefix-disk.img for qemu.

If the disk at <path>, or the default if -disk option is not provided, does not exist, linuxkit will create one of size <size>.

TODO: GCP

Mount the Disk

A disk created or used via hyperkit run will be available inside the image at /dev/vda with the first partition at /dev/vda1.

In order to use the disk, you need to do several steps to make it available:

  1. Create a partition table if it does not have one.
  2. Create a filesystem if it does not have one.
  3. fsck the filesystem.
  4. Mount it.

To simplify the process, two onboot images are available for you to use:

  1. format, which:
    • checks for a partition table and creates one if necessary
    • checks for a filesystem on the partition and creates one if necessary
    • runs fsck on the filesystem
  2. mount which mounts the filesystem to a provided path
onboot:
  - name: format
    image: "linuxkit/format:fdbfda789fe30a97ff194a06ac51ee0ff6b3ccf4"
    binds:
     - /dev:/dev
    capabilities:
     - CAP_SYS_ADMIN
     - CAP_MKNOD
  - name: mount
    image: "linuxkit/mount:ad138d252798d9d0d6779f7f4d35b7fbcbbeefb9"
    binds:
     - /dev:/dev
     - /var:/var:rshared,rbind
    capabilities:
     - CAP_SYS_ADMIN
    rootfsPropagation: shared
    command: ["/mount.sh", "/var/external"]

Notice several key points:

  1. format container
    • The format container needs to have bind mounts for /dev
    • The format container needs CAP_SYS_ADMIN and CAP_MKNOD capabilities
    • The format container only needs to run once, not matter how many external disks or partitions are provided. It finds all block devices under /dev and processes them.
  2. mount container
    • The mount container command is mount.sh followed by the desired mount point. Remember that nearly everything in a linuxkit image is read-only except under /var, so mount it there.
    • The mount container needs to have bind mounts for /dev and /var
    • The mount container needs CAP_SYS_ADMIN capabilities
    • The mount container needs rootfsPropagation: shared

With the above in place, if run with the current disk options, the image will make the external disk available as /dev/vda1 and mount it at /var/external.