- AWS CLI 1.10+
- AWS Account (to store artifacts, secrets)
- jabba for managing Java versions
- Rollbar Account (error reporting -- optional)
tl;dr:
export AWS_PROFILE=raster-foundry
export RF_SETTINGS_BUCKET=...
jabba use
-- if you don't have that version available, alsojabba install
the version from.jabbarc
./scripts/bootstrap
./scripts/update
./scripts/server
There are a set of tasks necessary before starting development in order to provision Raster Foundry. Raster Foundry depends heavily on AWS resources and using AWS resources to manage secrets/containers/artifacts in development. If only local development is being done, the primary resource that will be used are S3 buckets to store secrets.
In the AWS account you need to create a few buckets for the following:
- A bucket to house raw data (e.g. geotiffs, JPEG2000, ingest definitions, etc.)
- A config bucket that will store secrets for development and or other environments, an exported database for development data
- A bucket to house processed data (e.g. thumbnails, processed raster RDDs)
The names of the buckets are not important, but they should be memorable and easy to parse for your own sake. On your host machine you need to set up an AWS profile for the account with the S3 buckets. For instance, to set up an AWS profile called raster-foundry
with the AWS cli the following command would be used:
$ aws configure --profile raster-foundry
You will be prompted for an access key and secret key.
The .env.template
file is a template file with environment variables that get injected into running containers during development. This file should be copied into the AWS config bucket created after filling in sensitive information (replacing all PLACEHOLDER
values with appropriate values for your AWS setup). When provisioning this file is copied to the development environment and injected into containers with docker-compose
.
In addition to setting up an AWS account, you must register for an Auth0 account to produce secrets to use in the .env
file. You need to go through setting up an application and copying over the client IDs, domain, and secret.
Additionally, if you want to exercise token management in the application, you need to generate a management API app to handle managing the generation of refresh tokens for users via the management API. This is not necessary for most functionality in the application and can be deferred until later if you desire.
The last thing to set up with Auth0 are the allowed callback URLs and logout URLs. These need to be edited to allow interaction for local development from localhost:9091
and localhost:9100
.
Raster Foundry follows the approach outlined here ("Scripts to Rule Them All") to have a mostly consistent development experience. We deviate in a few specific ways:
- We don't pin / require a specific Java version. The application will eventually run in a jdk8 container, and for reproduction it's helpful to have
jabba
to be able to describe issues that occur on some Java versions but not others, but largely this does not make a difference at this point. - We expect the user to install
jabba
on their host, instead of running everything in containers. Users can choose to run everything in containers, but that's not how the development environment is organized by default.
Almost all interaction with consoles and servers can be managed via calls to a script located in ./scripts
. Default values for the S3 config and data buckets in addition to AWS profile will be used if they are not set with an environment variable. Before running scripts/bootstrap
, these should be injected into your shell environment:
export RF_AWS_PROFILE=raster-foundry
export RF_SETTINGS_BUCKET=rasterfoundry-development-config-us-east-1
After exporting your environment settings, you are ready to get started:
$ ./scripts/bootstrap
$ ./scripts/update
$ ./scripts/server
The servers should come up successfully.
Then, kill your servers. To get the database loaded with sample data, you can run ./scripts/load_development_data --download
.
This will fetch a database dump from S3 and some development images. You can use these data for consistent testing instructions
with other developers. This script will also apply any outstanding migrations not present in the dev database.
Database migrations are managed using flyway. You can run flyway
commands with scripts/migrate
. Some commands you
can run are:
scripts/migrate migrate
: apply outtanding migrationsscripts/migrate repair
: reconcile the checksums of applied migrations in the database with what's present on disk
There is no command to revert migrations.
The workflow for creating a new migration is:
- Write a migration in
db/src/main/resources/Vxx__migration_name.sql
./scripts/migrate migrate
- Verify the schema changes in PostgreSQL with
./scripts/psql
The Vagrant configuration maps the following host ports to services running in the virtual machines. Ports can be overridden for individual developers using environment variables
Service | Port | Environment Variable |
---|---|---|
Nginx (api) | 9100 |
RF_PORT_9100 |
Application Server (akka) | 9000 |
RF_PORT_9000 |
Tile Server (http4s) | 8081 |
RF_PORT_8081 |
Application Server (JMX) | 9010 |
RF_PORT_9010 |
Tile Server (JMX) | 9030 |
RF_PORT_9030 |
Helper and development scripts are located in the ./scripts
directory at the root of this project. These scripts are designed to encapsulate and perform commonly used actions such as starting a development server, accessing a development console, or running tests.
Script Name | Purpose |
---|---|
bootstrap |
Pulls/builds necessary containers |
update |
Runs migrations, installs dependencies, etc. |
server |
Starts a development server |
console |
Gives access to a running container via docker-compose run |
psql |
Drops you into a psql console. |
test |
Runs tests and linters for project |
cibuild |
Invoked by CI server and makes use of test . |
cipublish |
Publish container images to container image repositories. |
load_development_data |
Load data for development purposes from S3 |
rsync-back |
Perform a one-way rsync from the VM to the host. |
process-upload |
Process an upload in development |
Run all the tests:
$ ./scripts/test
In staging and production, a batch job will automatically be kicked off for processing after a successful upload. In development, you need to process the upload manually which you can do like so:
$ ./scripts/process-upload <upload_id>