atlasserver

https://fallingstar-data.com/forcedphot/

This is the source code for the ATLAS Forced Photometry Server, a Python Django Rest Framework server with a React frontend.

The ATLAS forced photometry server provides public access to photometric measurements over the full history of the ATLAS survey. After registration, a user can request forced photometry at any position on the sky either for a single position or a list of positions.

Why is the source code available?

The code is available for educational purposes, identification of security issues, and for curious users of the ATLAS Forced Photometry service. Pull requests are also welcome.

For ATLAS server administrators

The package should be installed in develop mode from the Git repository.

git clone https://github.com/lukeshingles/atlasserver.git
python3 -m pip install -e .

Copy dotenv_example.txt to .env and edit the relevant sections.

cp dotenv_example.txt .env

Then download the geoip database files with your MaxMind API key (stored in .env).

./update_geoipdatabase.sh

A MySQL or MariaDB server and tmux are required. These can be installed with homebrew on macOS:

brew install mariadb tmux

To initialise a new database, apply the migrations:

./manage.py migrate

Two processes must be running: the web server and the task runner. These can be started with:

atlaswebserver start
atlastaskrunner start

For atlastaskrunner to process jobs, there much be an SSH host alias named named 'atlas' that points to atlas-base-sc01.ifa.hawaii.edu with your username. The server-side scripts must also be installed in your sc01 home folder:

scp atlasserver/taskrunner/atlas_*.py atlas:~/

To update the code to the latest commit on the main branch, pull from the GitHub remote, apply any database changes, and then restart the two processes.

git pull
./manage.py migrate
atlaswebserver restart
atlastaskrunner restart

migrate is not optional: a pulled commit may change a model, and until its migration is applied every query against that table fails with an "Unknown column" error. It is safe to run when there is nothing to do.

Do not run makemigrations on the server. Migrations are committed to this repository, so the files that arrive with git pull are the ones to apply. Generating them on the server instead would produce files that differ from the ones under review, and — because a field rename is only detected by an interactive prompt that a non-interactive run answers "no" — a renamed field would be dropped and recreated empty rather than renamed.

Changing a model

Generate the migration on your own machine and commit it alongside the model change:

./manage.py makemigrations

Commit the generated file along with the model change. Django writes migrations in its own style, so ruff format will reformat them on commit; that is cosmetic and does not affect what the migration does. CI fails if a model change arrives without its migration.

License

Copyright (c) 2020-2024 Luke Shingles
Distributed under the MIT license. See LICENSE for more information.