Skip to content

Backend service to manage pipeline projects and processing.

License

Notifications You must be signed in to change notification settings

MouseLightPipeline/pipeline-api

Repository files navigation

Pipeline Coordinator Service

This service is the Mouse Light Acquisition Pipeline Coordinator. It manages pipeline projects, pipeline stages for those projects, and the task definitions that define units of work for the stages.

Dependencies

The Pipeline Database service must be running to synchronize task definitions between this service and any workers.

Installation

The service can be run standalone or within a Docker container.

In either case there four environment variables that can be used to override the default values for connecting to the pipeline database server. These also apply to the migrations and seeding sections below.

  • PIPELINE_DATABASE_HOST (default "pipeline-db")
  • PIPELINE_DATABASE_PORT (default 5432)
  • PIPELINE_DATABASE_USER (default postgres)
  • PIPELINE_DATABASE_PASS (default pgsecret)

When using the container on the same host as the database container the default values can be used.

When running standalone and the system entirely on localhost (e.g., during development), or when running the container distributed from the database container, at a minimum PIPELINE_DATABASE_HOST must be set.

For production the postgres user/pass should of course be changed.

Standalone

As a standalone service it requires node 7.10 or later.

Outside of a container the service can be started in a number of ways

  • run.sh ensures that all migrations are up to date and uses nohup to facilitate disconnecting from the session
  • npm run start or npm run devel for production or development mode (node debugger attachment and debug messages) without migration checks
  • docker-entry.sh ensures that all migrations are up to date and launches the node process and sleeps

run.sh is useful for starting from an interactive session where you plan to disconnect.

docker-entry.sh may be useful where nohup is not necessary and/or you want the script to stay alive. For example, this is used by the container where the sleep is required to keep the container active.

Container

As a container the service only requires that Docker be installed. The container can be started via docker-compose (see the server-prod repository as an example) or directly from the docker command.

The container has the same startup options as the standalone via the CMD compose property or docker run command line argument, but defaults to docker-entry.sh.

The offline database is stored in /app/internal-data. To persist offline data between container updates, map this directory to a host volume or data volume. This is required as the offline database only synchronizes task information from the coordinator database at this time - it does not sync project and stage information.

Migrations

Database migrations for both the offline and coordinator database can be run manually using migrate.sh. This can be done for a standalone instance directly, or via docker run as the command.

By default the coordinator database configuration assumes a host of pipeline-db and port 5432. To migrate localhost or a different remote database and or use a different port, pass the host and port as argument to the script or set the PIPELINE_DATABASE_HOST and/or PIPELINE_DATABASE_PORT environment variables.

Seeding

There are default seed data sets with sample projects and tasks for "development" and "production" environments.

By default the coordinator database configuration assumes a host of pipeline-db and port 5432. To migrate any other configuration, pass the host and port as arguments to the script or set the PIPELINE_DATABASE_HOST an PIPELINE_DATABASE_PORT environment variables directly.

The default seed is the production seed. To use the development seed, pass "development" as the third argument to the script, or set the PIPELINE_SEED_ENV environment variable directly. Note that if you use as the third argument you must pass the correct/valid host and port information as the first two arguments, even if environment variables are set.

Example Installations

The simplest method to be up and running is using Docker Compose and the services-prod repository configuration. This will launch this service, the graphical front end server for this service, and all supporting services such as the pipeline database.

Local Development

The following assumes that the pipeline database is run via the services-prod repo Docker Compose configuration.

You can also run a postgres server anywhere/way that contains a pipeline_production database, but you must change any values for the pipeline database host and port from localhost and 4432 below to their appropriate values if different.

  1. Start the pipeline database using the server-prod repo and the up.sh script to start all background services and data volumes via Docker Compose.
  2. Stop this api service container via docker stop <container_id>. You can find the id for the pipeline_api container using docker ps
  3. If this is the first launch, or if a new migration has been added, use the migrate.sh script as ./migrate.sh localhost 3932.
  4. If this is the first launch, optionally seed the database to get up and running quickly using ./seed.sh localhost 3932
  5. Launch using npm run devel

Expected Input

Input Source

The pipeline first attempts to read from a file named pipeline-input.json in the project root directory. If pipeline-input.json does not exist, the system will look for dashboard.json. These files can be in either of the following formats.

Dashboard Input Format

The pipeline can read the standard MouseLight Acquisition Dashboard output file - dashboard.json. This format is documented elsewhere.

Pipeline Input Format

This pipeline also supports a streamlined version of the dashboard.json data with only the elements required for pipeline processing. This file can be generated outside of the MouseLight acquisition process and for data other than MouseLight acquisitions.

The format the following required structure:

{
  "tiles":
     [
       {
         "id": 1,
         "relative_path": "2017-12-20\\00\\00001",
         "isComplete": true,
         "position": {
           "x": 197,
           "y": 119,
           "z": 1867
         },      
         "step": {
           "x": 197,
           "y": 119,
           "z": 1867
         },
       },
       {
         "id": 2,
         ...
       }
     ],
     "1":
     [
      ...
     ]
}

The tiles property is an array of tile information. The required components, structure, and data types are shown above. id is an optional numeric value. relativePath can be Windows or Unix format and will be normalized to Unix for internal use. In either case it must be escaped properly, as shown above, to be interpreted correctly. The value is a relative path from the project root directory.

The isComplete property is used to determine whether to pass the tile to any stage 1. If false, it will appear as a known tile in the Tile Heat Map display page, known tile counts for the project, etc..., but will not be processed.

Project Information

There is a top-level section of the file that can be used to provide information about the project itself.

The pipelineFormat field is required to indicate it is not a dashboard input file.

The extents field defines the extents of the sample and is optional, but if present should be a valid range for the sample. This improves/simplifies some of the information displayed about the project, but is not required.

The section is at the same level as the tiles property and is as follows:

{
    "pipelineFormat": "default",
    "extents":
    {
      "minimumX": 184,
      "maximumX": 214,
      "minimumY": 115,
      "maximumY": 130,
      "minimumZ": 1867,
      "maximumZ": 1941
    },
    "tiles":{
    ...
    }
}

About

Backend service to manage pipeline projects and processing.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages