Update Deployment documentation

[phycodurus]

The documentation page, "Deploy a TOM to Heroku" is out of date. Heroku discontinued their free-tier in November, 2022.

The deployment section should probably be re-written to emphasize writing a Dockerfile and locally testing a dockerized TOM because this is probably a prerequisite to most deployment strategies. In the meantime, tom-demo has a Dockerfile that can be used as a starting off point.

This should include instructions on how to incorporate whitenoise middleware for handling static files since this will become an issue when running from a Dockerfile.

[gnthibault]

Anyone deployed tom in google cloud platform ? I saw django can be configured to serve static files from google cloud storage using S3 sample but with gcs like here

pip install django-storages[google] and set: DEFAULT_FILE_STORAGE = "storages.backends.gcloud.GoogleCloudStorage" STATICFILES_STORAGE = "storages.backends.gcloud.GoogleCloudStorage" GS_BUCKET_NAME = 'your_bucket_name' STATIC_URL = 'https://storage.googleapis.com//' from google.oauth2 import service_account GS_CREDENTIALS = service_account.Credentials.from_service_account_file( 'path/to/the/downloaded/json/key/credentials.json' # see step 3 )

Those previous parameters, are actually better to be defined in a google cloud secret manager, and imported at install-time by changing the settings.py see:

• https://cloud.google.com/python/django/run#understanding-secrets
• https://cloud.google.com/python/django/run#database_connection
• https://cloud.google.com/python/django/run#cloud-stored_static

Equivalent cors can be defined from the bucket directly, with an IAC tool like openTOFU or terraform like in here: https://registry.terraform.io/providers/hashicorp/google/latest/docs/resources/storage_bucket

I am still working on GCP deployment, will update if I manage to go through, at the moment, I am reading this overview of the various options from GCP: https://cloud.google.com/python/django

Edit, I am most likely going for a cloudrun deployment, as it can cope with explicit docker images, and can scale to zero instance, which is good in my case: https://cloud.google.com/python/django/run and https://codelabs.developers.google.com/codelabs/cloud-run-django#0

I am also using the manual install guide: https://tom-toolkit.readthedocs.io/en/latest/introduction/manual_installation.html and as suggested by @phycodurus the dockerfile from tom_demo: https://github.com/LCOGT/tom-demo/blob/57ae7565f8e1f2864a2478ccfe1f11a696e4c067/Dockerfile But also the one suggested in the documentation ie dockertom (not maintained either unfortunately) https://github.com/TOMToolkit/dockertom/blob/907b98fe826c44ce3ea79148503ef2d7ac8fd4fc/Dockerfile Or yet another one from panoptes project: https://github.com/gnthibault/panoptes-tom/blob/b676abd8ae8a6b6871f4a922e815ce2f1903c022/Dockerfile

Ok, I also see this advice from https://tom-toolkit.readthedocs.io/en/1.2.6/customization/automation.html that doesn't seems to be in the demo dockerfile:

crontab should be as follows: 0 /path/to/virtualenv/bin/python /path/to/project/manage.py save_data

EDIT: ok I start to have serious doubt about cloudrun, see the actual container contract:

For Cloud Run jobs, the container must exit with exit code 0 when the job has successfully completed, and exit with a non-zero exit code when the job has failed.

Because jobs should not serve requests, the container should not listen on a port or start a web server.

EDIT2: actually, it seems perfectly fine, see https://cloud.google.com/run/docs/tips/python: CMD exec gunicorn --bind :$PORT --workers 1 --threads 8 --timeout 0 main:app

Additional django docker samples: https://github.com/docker/awesome-compose/tree/master/django https://github.com/nickjj/docker-django-example