CARDS (Clinical Archive for Data Science)

Based on Apache Sling

Prerequisites:

Java 1.8+
Maven 3.3+
Python 2.5+ or Python 3.0+
psutil Python module (recommended)

Build:

mvn clean install

Additional options include:

mvn clean -Pclean-node to remove compiled frontend code

mvn clean -Pclean-instance to remove the sling/ folder generated by running Sling

mvn install -Pquick to skip many of the test steps

mvn install -Pskip-webpack to skip reinstalling webpack/yarn and dependencies and skip regenerating the frontend with webpack

mvn install -PautoInstallBundle to inject compiled code into a running instance of Sling at http://localhost:8080/ with admin:admin.

To specify a different password, use -Dsling.password=newPassword

To specify a different URL, use -Dsling.url=https://cards.server:8443/system/console (the URL must end with /system/console to work properly)

mvn install -PintegrationTests to run integration tests

Run:

./start_cards.sh => the app will run at http://localhost:8080 (default port)

./start_cards.sh -p PORT to run at a different port

./start_cards.sh -Dsling.run.modes=dev to include the content browser (Composum), accessible at http://localhost:8080/bin/browser.html

Installing larger vocabuleries may fail due to default limits imposed on XML documents by the JVM. In this case, the app should be started with an extra parameter:

./start_cards.sh -Djdk.xml.entityExpansionLimit=0

By default, the app will run with username admin and password admin.

In order to use "Vocabularies" section and load vocabularies from BioPortal (bioontology.org) BIOPORTAL_APIKEY environment variable should be set to a valid BioPortal API key. You can request a new account if you don't already have one, and the API key can be found in your profile.

Running with Docker

If Docker is installed, then the build will also create a new image named cards/cards:latest

Test/Development Environments

CARDS can be ran as a single Docker container using the file system (instead of Mongo) as a data storage back-end for Apache Sling.

docker run --rm -e INITIAL_SLING_NODE=true -e OAK_FILESYSTEM=true -p 127.0.0.1:8080:8080 -it cards/cards

Production Environments

Before the Docker container can be started, an isolated network providing MongoDB must be established. To do so:

docker network create cardsbridge
docker run --rm --network cardsbridge --name mongo -d mongo

For basic testing of the CARDS Docker image, run:

docker run --rm --network cardsbridge -e INITIAL_SLING_NODE=true -d -p 8080:8080 cards/cards

However, since runtime data isn't persisted after the container stops, no changes will be permanently persisted this way. It is recommended to first create a permanent volume that can be reused between different image instantiations, and different image versions.

docker volume create --label server=production cards-production-volume

Then the container can be started with:

docker container run --rm --network cardsbridge -e INITIAL_SLING_NODE=true --detach --volume cards-test-volume:/opt/cards/sling/ -p 8080:8080 --name cards-production cards/cards

Explanation:

docker container run creates and starts a new container
--rm will automatically remove the container after it is stopped
--network cardsbridge causes the container to connect to the network providing MongoDB
--detach starts the container in the background
-e INITIAL_SLING_NODE=true marks this container as the first to start up, and thus responsible for setting up the database
--volume cards-test-volume:/opt/cards/sling/ mounts the volume named cards-test-volume at /opt/cards/sling/, where the application data is stored
-p 8080:8080 makes the local port 8080 forward to the 8080 port inside the container
- you can also specify a specific local network, and a different local port, for example -p 127.0.0.1:9999:8080
- the second port must be 8080
--name cards-production gives a name to the container, for easy identification
cards/cards is the name of the image

To enable developer mode, also add --env DEV=true -p 5005:5005 to the docker run command.

To enable debug mode, also add --env DEBUG=true to the docker run command. Note that the application will not start until a debugger is actually attached to the process on port 5005.

docker run --network cardsbridge -d -p 8080:8080 -p 5005:5005 -e INITIAL_SLING_NODE=true --env DEV=true --env DEBUG=true --name cards-debug cards/cards

Running with Docker-Compose

Docker-Compose can be employed to create a cluster of N MongoDB Shards, M MongoDB Replicas, and one CARDS instance.

Installing/Starting

Before proceeding, ensure that the cards/cards Docker image has been built.

mvn clean install

The ccmsk/neuralcr image is also required. Please build it based on the instructions available at https://github.com/ccmbioinfo/NeuralCR. Use the develop branch.

Download the pre-trained NCR models from here and un-tar. Create the directory NCR_MODEL under compose-cluster and copy in the file pmc_model_new.bin along with the directories 0 and 1 from the ncr_model_params directory.

Now build the docker-compose environment.

cd compose-cluster
python3 generate_compose_yaml.py --shards 2 --replicas 3
docker-compose build

3.1 Replacing python3 generate_compose_yaml.py --shards 2 --replicas 3 with python3 generate_compose_yaml.py --oak_filesystem will not start a Mongo cluster and instead will use the file system as a data storage back-end for Apache Sling

Start the docker-compose environment.

docker-compose up -d

The CARDS instance should be available at http://localhost:8080/

5.1. To inspect the data split between the MongoDB shards:

docker-compose exec router mongo
sh.status()
exit

Stopping gracefully, without losing data

To stop the MongoDB/CARDS cluster:

docker-compose down

Restarting

To restart the MongoDB/CARDS cluster while preserving the entered data from the previous execution:

CARDS_RELOAD=true docker-compose up -d

Cleaning up

To stop the MongoDB/CARDS cluster and delete all entered data:

docker-compose down #Stop all containers
docker-compose rm #Remove all stopped containers
docker volume prune -f #Remove all stored data
./cleanup.sh #Remove the cluster configuration files

sdumitriu/cards