User guide for HYLODE

HySys

Your first port of call for all questions HySys should be the top-level repository documentation. Some tips and tricks for particularly common patterns are given here.

Deployment

Deployment instructions for a fresh installation of HySys are provided in the readme linked to above. To redeploy an already existing deployment on the NHS network:

  1. Navigate to the directory containing the checked out repository. Locations are provided here for the staging_red and prod deployments.
  2. Navigate to the the hylode/ directory (containing docker-compose.yml)
  3. From within a tmux or screen session, execute the bin/up.sh script to deploy most of the back-end HySys services
    • you can execute this by running ./bin/up.sh on the command line
    • it’s recommended to run this command inside a tmux or screen session

This will redeploy all non-interactive services - everything but the HyMind Lab and Field.

Interactive service deployment

To deploy the HyMind Lab and HyMind Field services, execute ./bin/interactive-up.sh. Again, it’s recommended that you do this within a screen or tmux session.

Taking services down

To take down all running services - including interactive services described above:

  1. Navigate to the directory containing the checked out repository. Locations are provided here for the staging_red and prod deployments.
  2. Navigate to the the hylode/ directory (containing docker-compose.yml)
  3. Run ./bin/down.sh from the command line
    • Optionally provide the --remove-orphans flag to remove hanging containers and networks. This will not remove any associated docker volumes, so is safe from a data-loss perspective.

In-band predictions

Current and past model predictions arising as a result of the HySys pipeline are written to corresponding tables within the hymind schema within the hylode database found in the postgres service of each environment.

The latest predictions are available via the HyMind API REST endpoint. The HySys documentation details the URL of the documentation of this API - look for HYMIND_API_WEB_PORT.

Please note

  • the application is only accessible from within the UCLH network
  • usernames and passwords are available after training from the development team