This is a overlayFS manager that provides 'stacking' functionalities similar to Docker images, but not tied specifically to a container solution. With Stacko you can create immutable "images" which stack together. You can create one or more writable "instances" of a stack that have a thin Copy-On-Write layer. These instances are usable at a named "stackpoint", which always points to an instance. Eventually you can cutover a stackpoint to a new instance and fallback to a previous instance without copying or destroying files.
Note: this is a work in progress, so it is not stable or finished yet.
This was a weekend project, which was successful! However, it appears there is a more
promising alternative to Stacko named Siren: https://github.com/LEW21/siren .
This package requires the following packages to function:
- https://github.com/wagoodman/overlayUtils
- subwrap (pip install subwrap)
A working concept of how Stacko works:
# create and populate a new image 'foundation-libs0.1'
stacko new-image foundation-libs0.1
stacko edit-image foundation-libs0.1
...<head to the mount directory and add/edit contents>
stacko close-image foundation-libs0.1
# create another image 'apps-1.0' which "stacks" on image 'foundation-libs0.1'
stacko new-image apps-1.0 foundation-libs0.1
stacko edit-image apps-1.0
...<head to the mount directory and add/edit contents>
stacko close-image apps-1.0
# create a stackpoint named 'system1fs'. This automatically creates a new instance
# of the 'apps-1.0' image
stacko new-stackpoint system1fs apps-1.0
# mount the 'system1fs' for usage...
stacko mount-stackpoint system1fs
<reboot>
# This is all you need to do to use this stack again!
stacko mount-stackpoint system1fs
...
# Time to upgrade to another version of the application
stacko new-image apps-2.0 foundation-libs0.1
<populate it as before>...
# Use this new image as a new instance for the 'system1fs' stackpoint:
stacko cutover-stackpoint system1fs apps-2.0
# Oh no! something is wrong and you need to go back to the previous instance used
# (this includes all changes made in the CoW layer)
stacko fallback-stackpoint system1fs
This came about when looking for a solution that provided the prescriptive-ness of
a Docker image, something that can be stacked and referenced, but not written to.
Though, using this solution implies that you will use Docker --the image/layer/storage
solution is closely coupled to the container engine. This means that you need to
run the Docker daemon to create images, even if you intend to export them for use
with systemd-nspawn instead. Secondly, exporting the image merges all of the layers
into a single directory! Though docker save keeps the layers in separate directories
you are now closely tied to the Docker image format.
What I really wanted (right or wrong) was the ability to manage this concept of immutable "images" in layers, with the ability to fork off CoW instances on a whim, and do all of this without being tied to a particular container engine.
usage: stacko <command> [<args>]
Image commands:
new-image Create an image
edit-image Mount an image for editing
close-image Umount an image to stop editing
delete-image
list-images Show the existing images
import-image
export-image
StackPoint commands:
new-stackpoint
delete-stackpoint
cutover-stackpoint-instance: unmount a current instance, create a new
instance of the given image (if it does not already exist), set the
current instance, and mount the new current instance
fallback-stackpoint-instance: unmount a current instance, set the
current instance to the last known image instance, and mount the
new current instance
new-stackpoint-instance
set-stackpoint-instance
delete-stackpoint-instance
mount-stackpoint
umount-stackpoint
get-stackpoint-dir
is-stackpoint-mounted