This brings together project https://github.com/Avature/lxd-image-server (well, from my fork, https://github.com/cr1st1p/lxd-image-server, branch import-metadata , since it contains various fixes + improvements) + a file uploader server, all, in one.
It creates a docker image out of those parts.
You should run that image behind a SSL terminating proxy - LXD for example accepts only https simplestream repositories. For example, run this inside a Kubernetes cluster.
You will need to mount a volume into directory /var/www/simplestreams
Simplestream is a 'protocol' (description) for describing what virtual machine images exist on a http(s) repository.
It mainly consists on serving, besides the actual machine images content, 2 files:
- streams/v1/index.json
- streams/v1/images.json
images.json contains a list of all the virtual machine images, with various information about them.
What the lxd-image-server does (but it is not related strictly to lxd), is to automatically update the images.json file when it detects new files uploaded. Well, it does some external repositories updates as well, but that's a different story.
This image also adds a simple web server file upload feature, so it is more of an integrated solution.
You should protect access to this server (via the SSL terminating proxy)
Please check the section This is the structure the simplestreams server needs to have. at https://github.com/Avature/lxd-image-server/blob/master/README.md to understand how the images you upload should be named and in which directories they should end.
Code will set various fields based on that directory layout! Changes in the particular forked repository also allows you to upload a metadata.json with which you can override:
- os
- release_title
- aliases (an array)
As usual, via environment variables:
- DEBUG : will make programs output more things
- ALLOW_OVERWRITES: if non empty, it will allow you to update files, else, it will create them with a changed filename
- ALLOW_DELETES: if non empty, it allows you to delete files via http method DELETE
Please configure your front proxy for big enough body sizes since you're uploading virtual images... You can see such a setting, in example/.yaml*, where we set nginx ingress controller to accept a body size of 10G.
I strongly recommend adding authorization in that same front proxy layer - the file upload code is not battle tested so you should limit access to at least that part.
lxc remote add my_remote https://...your_server --protocol=simplestreams
lxc image list my_remote:Obtain the image(s). Example with lxd:
mkdir -p /tmp/img
cd /tmp/img
lxc image list
lxc image export 9e7158fc0683
ls -lWe have 2 files (in this case - but if you export an image obtained from a container, things are more complicated): 9e7158fc0683d41......squashfs and meta-9e7158fc0683d41.....tar.xz (digits removed)
This is an ubuntu image, architecture amd64.
product=ubuntu
release=xenial
arch=amd64
box=default
remote=https://your_install.com
# you MUST obey the date format
upload_date=$(date '+%Y%m%d_%H:%M')
dir="$product/$release/$arch/$box/$upload_date"
# HERE, you might first want to edit a metadata.json and upload it
# ...
curl "[email protected];filename=$dir/lxd.tar.xz" "$remote/images/"
curl "[email protected];filename=$dir/rootfs.squashfs" "$remote/images/"
# check, after seconds (code needs to compute checksums), the presence in:
curl -v "$remote/streams/v1/images.json"
# or, via lxc client:
lxc image list my_remote:For an image created from a container, when you do an lxc image export fingerprint you will obtain a single TAR file. This one contains both metadata and the image content itself. You'll have to manually create an lxd.tar.xz file out of it, without the rootfs/ directory, and a root.tar.xz with the content from rootfs/
A sample yaml file can be found in kubernetes/ directory.
It is not functional without editing some things in it, like:
- namespace field
- domain name
- ingress secret ref and secret name
- persistent storage class name