Song

Quickly and reliably track genome metadata scattered across multiple Cloud storage systems.

Project containing both the SONG microservice and CLI client. Both are written using JAVA 8 and Spring Boot.

Documentation

Explore documentation with the Song Read the Docs.

Build

$ mvn clean package

Running

Command-line

The source can be built and run using maven.

$ git clone [email protected]:icgc-dcc/SONG.git
$ cd SONG/song-server
$ mvn spring-boot:run -Dspring-boot.run.profiles=dev,test

Both the server and client when compiled and built produce uber jars which can be run easily from the command line.

$ java -jar song-server-0.1.1-SNAPSHOT.jar  --spring.profiles.active=dev,test

Docker Based Developement

Several make targets are provided for locally deploying dependent services using docker. By using this, the developer will be able to replicate a live environment for song-server and song-client. It allows the user to develop locally, and test submissions, manifest creation, publishing, unpublishing and score uploads/downloads in an isolated environment.

There are 2 modes:

1. Developement Mode

The purpose of this mode is to decrease the wait time between building and testing against dependent services. This mode will run a mvn package if the distribution files are missing and copy them into a container for them to be run. This method allows for fast developement, since the mvn package step is handled on the Docker host. In addition, the debug ports 5005 and 5006 are exposed for both song-client and song-server, respectively, allowing developers to debug the docker containers. This mode can be enabled using the DEMO_MODE=0 override. This is the default behaviour if the variable DEMO_MODE is not defined.

Debugging the song-client with IntelliJ

Since the JVM debug port is exposed by the song-client docker container, IntelliJ can remotely debug a running docker container. To do this, a docker image run profile must be created with the configuration outputted by the make intellij-song-client-config command, which will output a basic upload command, however it can be modified to be any song-client command. Then, a remote debug profile must be created, with the following config:

Host: localhost
Port: 5005
Use module classpath: song-client

and in the Before launch: Activate tool window section, click the + sign, and select Launch docker before debug. Then ensure the Docker configuration field is set to the name of the previously created docker image run profile and that Custom Options is set to -p 5005:5005. In order for the debugger to bind to the debug port in time, a delay needs to be introduced after starting the container. To do this, click the + sign again, and select Launch docker before debug, and select Run External Tool and a window will pop-up. Input the following:

Name:      Sleep for 5 seconds
Program:   /usr/bin/sleep
Arguments: 5

and click OK.

Finally, start debugging by simply running the remote debug profile and it will call the docker image run profile before launch.

Executing the dockerized song-client in developement mode

The script ./docker/tools/song-client-dev takes the arguments runs the song-client service entry specified in the docker-compose.yml with them. For example, to ping the song server, run ./docker/tools/song-client ping

Debugging the song-server with IntelliJ

Since the song-server is a server and exposes the 5006 debug port, configuration is much easier. First, start the server with make clean start-song-server. Then, create a remote debug profile in Intellij with the following configuration:

Host: localhost
Port: 5006
Use module classpath: song-server

and then run it in debug mode.

2. Demo Mode

The purpose of this mode is to demo the current song-server and song-client code by building it in inside the Docker image, as opposed to the Docker host as is done in Developement mode and then running the containers. This mode will not run mvn package on the Docker host, but instead inside the Docker container. This method is very slow, since maven will download dependencies every time a build is triggered, however creates a completely isolated environment for testing. This mode can be enabled using the DEMO_MODE=1 make variable override. For example, to start the song-server, the following command would be run:

Executing the dockerized song-client in demo mode

The script ./docker/tools/song-client-demo takes the arguments runs the song-client service entry specified in the docker-compose.yml with them. For example, to ping the song server, run ./docker/tools/song-client ping

make start-song-server DEMO_MODE=1

For more information on the different targets, run make help or read the comments above each target for a description

API

The server provides swagger docs documenting the API.

When running locally they can be accessed here: http://localhost:8080/swagger-ui.html

Dockerhub Configuration

1. Edit build configurations by selecting the Builds tab at the top, then click Build Configuration

2. Create a new build rule by clicking the + sign beside the BUILD RULES text

3. Edit the configuration as follows: Source Type: branch Source: develop Docker Tag: develop Dockerfile Location: Dockerfile Build Context: / Autobuild is set to the ON position

   Build Caching is set to the ON position

4. Then save the configuration

Docker Song Client

The song-client is a CLI tool used for communicating with a song-server.

Building

Simply running mvn clean package will package the client into a -dist.tar.gz file.

Configuration

After unarchiving the distribution, it can be configured via the ./conf/application.yml file. Alternatively, the client can be configured through environment variables, which take presedence over the application.yml config. For example, to run the song-client config command using environment variables with the same values as the application.yml configuration below:

client:
  serverUrl: http://localhost:8080
  studyId: ABC123-CA
  programName: sing
  debug: true
  accessToken: myAccessToken

could be done via:

CLIENT_SERVER_URL=http://localhost:8080 \
CLIENT_STUDY_ID=ABC123-CA \
CLIENT_PROGRAM_NAME=sing \
CLIENT_DEBUG=true \
CLIENT_ACCESS_TOKEN=myAccessToken \
./bin/sing config

Running the client locally

The song-client can be run using the ./bin/sing script.

Running the client using Docker

Alternatively, the song-client can be run using docker. To run the dockerized client with the configurations above, the following command could be executed:

docker run --rm \
  -e 'CLIENT_SERVER_URL=http://localhost:8080' \
  -e 'CLIENT_STUDY_ID=ABC123-CA' \
  -e 'CLIENT_PROGRAM_NAME=sing' \
  -e 'CLIENT_DEBUG=true' \
  -e 'CLIENT_ACCESS_TOKEN=myAccessToken' \
  overture/song-client:latest \
  sing config

By default, the song-client is run as the root user. To run as a non-root user, add the switch -u song which will run the command as a predefined song user:

docker run --rm \
  -u song \
  -e 'CLIENT_SERVER_URL=http://localhost:8080' \
  -e 'CLIENT_STUDY_ID=ABC123-CA' \
  -e 'CLIENT_PROGRAM_NAME=sing' \
  -e 'CLIENT_DEBUG=true' \
  -e 'CLIENT_ACCESS_TOKEN=myAccessToken' \
  overture/song-client:latest \
  sing config

or run it as your current user:

docker run --rm \
  -u $(id -u):$(id -g) \
  -e 'CLIENT_SERVER_URL=http://localhost:8080' \
  -e 'CLIENT_STUDY_ID=ABC123-CA' \
  -e 'CLIENT_PROGRAM_NAME=sing' \
  -e 'CLIENT_DEBUG=true' \
  -e 'CLIENT_ACCESS_TOKEN=myAccessToken' \
  overture/song-client:latest \
  sing config

Running as the host user is useful when the song-client needs to write to a mounted volume

Outputting data from the song-client via Docker

Some song-client commands (such as sing manifest and sing export) output a file to a path. When running the docker container, it maybe preferable to output the file to the docker host's filesystem, instead of the containers file system. To do this, a directory from the docker host must be mounted into the song-client docker container.

For example, the following command will generate a manifest file called output-manifest.txt in the directory ./mydir:

# Ensure the current user owns mydir inorder to write to it from within the container
mkdir -p ./mydir

docker run --rm \
  -u $(id -u):$(id -g) \
  -v $PWD/mydir:/data \
  -e 'CLIENT_SERVER_URL=http://localhost:8080' \
  -e 'CLIENT_STUDY_ID=ABC123-CA' \
  -e 'CLIENT_PROGRAM_NAME=sing' \
  -e 'CLIENT_DEBUG=true' \
  -e 'CLIENT_ACCESS_TOKEN=myAccessToken' \
  overture/song-client:latest \
  sing manifest -a someAnalysisId -d /data -f /data/output-manifest.txt

Notes

When running with the secure profile enabled, an oauth2 server is needed.

Test