create a locally hosted helm repository to use in dependencies, serve using python http.server

helmserver.py

def helmserver(dirname, hostname='localhost', port=8000):
    # runs a helmserver (http server serving index.yaml and helm packages)
    # returns a multiprocessing.Process, call .terminate() to stop it
    # Usage:
    #   helmserver('/path/to/dependencies', hostname, port)
    def httpdrunner(dirname, bind, port):
        # this adopted from http.server module
        import os
        server_address = (bind, port)
        os.chdir(dirname)
        httpd = HTTPServer(server_address, SimpleHTTPRequestHandler)
        httpd.serve_forever()

    p = Process(target=httpdrunner, args=(dirname, hostname, port))
    p.start()
    sleep(1)  # give it some time to start
    return p

Makefile

# put this in ./dependencies
# add external dependencies that cannot be added as a repository
# then add this directory as a file
# for every external chart, script the download as shown in localpath example, add to build target dependencies
# use $ make build to download all dependencies, build packages and local repo index

localpath:
	mkdir -p staging
	curl https://codeload.github.com/rancher/local-path-provisioner/tar.gz/v0.0.15 -o staging/localpath.tgz
	cd staging && tar -xzf localpath.tgz
	cd staging/local-path-provisioner*/deploy/chart && helm package .
	mv staging/local-path-provisioner*/deploy/chart/*.tgz .
	rm -rf staging

build: localpath
	helm repo index .

Why?

• sometimes your helm chart depends on another helm chart hosted in some github repo, but not served in a proper helm repo
• you don't really want to clone the repo and keep the full chart, you really just want to keep the charts tgz
• you want everything automated

Create a local helm repository of your externally hosted packages, e.g.

$ mkdir dependencies
$ cp Makefile dependencies
$ cd dependencies && make build
$ cd dependencies && helm repo index .

Specify in your chart dependencies:

# Chart.yaml
dependencies:
    - name: chart-name
      repository: http://localhost:8000/

Then serve your packages using before calling helm dependency update

$ cd dependencies && python -m http.server &
$ helm dependency update

If you have some python program that does all the helm'ing for you, use the above to run the http.server in a process instead of separately

# myprogram.py, e.g. invoke tasks
from helmserver import helmserver
@task
def update(c):
     p = helmserver('../dependencies')   
     c.run('helm dependency update .')
     p.terminate()