Usage example

As videos are large, often the videos are hosted on a remote machine that is different to the one you will manipulate metadata, and commit to git on.

So, these examples will have commands being executed on 2 machines, the “uploader machine” (e.g. apu for DebConf) and “your machine”. But you can, of course, do this all on a single machine.

The once-off setup below applies to the uploader machine.

Once-off setup

YouTube authentication: credential creation

You can skip this if someone gives you a client_secrets.json.

  1. Create a Google APIs project. https://console.developers.google.com/apis/

  1. Enable the YouTube Data v3 API in the “Library”.

  2. Create a “Desktop” OAuth 2.0 client ID in the “Credentials”.

  3. Download the Client ID JSON, and save it as client_secrets.json.

YouTube authentication: authorization

  1. Get a client_secrets.json from another team member, or generate your own, following the YouTube credential creation steps.

  2. Run auth.py (in the virtualenv, below). It will run a webserver on localhost:8000. Port-forward this to your local machine, if you are running it on a remote server. It will also attempt to open a URL in a web browser. If that isn’t possible, the URL is printed out for you to open by hand.

  3. Visit that URL, and sign in with a Google account that has admin access to the YouTube channel. Select the channel, and authorize the app to access it.

  4. You can now delete client_secrets.json, if you want. The uploader only needs client_id.json.

Virtualenv creation

This uploader requires a few Python libraries (see requirements.txt). You can either install them from Debian, or create a virtualenv to contain them.

  1. Create a new Python virtualenv:

    python3 -m venv ve
ve/bin/python -m ensurepip -U
ve/bin/python -m pip install wheel

  2. Install the required packages in the virtualenv:

    ve/bin/python -m pip install -r requirements.txt

  3. Activate the virtualenv before using these tools, every time:

    . ve/bin/activate

Import videos from a conference

  1. Upload the videos to the meetings-archive.

  2. Describe them in archive-meta.

  3. Run merge-meta:

    tools/merge-meta.py ../archive-meta/metadata/

  4. A new metadata file should be created with your new playlist. Check the contents, edit archive-meta if necessary, and repeat until satisfied with the data.

  5. Check and update the defaults, applied to every upload.

  6. Optional: Create a playlist in YouTube, and record this in the defaults.

  7. Copy this YouTube metadata file to the uploader machine with the meetings-archive available locally.

  8. Run the upload script, on the uploader machine, in the base directory for the conference videos, recording uploaded video IDs to a log.

    The first time you do this, you’ll have to authorize it.

    cd /srv/video/video.debian.net/2019/example/
~/youtube/tools/upload_youtube.py --wait \
    ~/youtube/metadata/2019/example.yml ~/upload.log

  9. Copy upload.log back to your machine.

  10. Merge the video IDs from it into the metadata file:

    tools/merge-uploads.py metadata/2019/example.yml upload.log

  11. The --wait will cause the uploader to wait and retry, until the entire playlist is uploaded, even if YouTube rate-limits are reached.

    If you aren’t using --wait, or have another failure, and need to continue a partial upload:

    1. Ensure the uploaded video IDs have been merged into the metadata (from the above steps).

    2. Update the metadata on the uploader machine.

    3. Run the upload script again. It will upload videos without video_id attributes.