Skip to content

Deployment

The project is deployed on CERN's PaaS platform as two separate deployments:

  • mlplayground using a modified s2i deployment,
  • dqm-playground-ds from a Docker image.

mlplayground

Due to the root dependency that opening nanoDQM files introduces, a custom s2i base image has been created using the procedure followed here.

See the Dockerfile for the extra packages added to the default RHEL UBI8 image.

Running management commands on OpenShift

Due to the project's dependency on CERN's ROOT, in order to run any management command on OpenShift (i.e. discover_dqm_files) you must run source /opt/app-root/src/root/bin/thisroot.sh to add ROOT to PATH.

dqm-playground-ds

This project is automatically built into a Docker image using GitHub Actions and published on Docker Hub.

On each new image build, an update is rolled out and the deployment is updated.

This was configured using these instructions.

PaaS Resource Limits

Histogram parsing from .csv files can be pretty CPU-intensive, which the default PaaS limits cannot handle.

Follow the guide to increase the limits to:

resources:
    limits:
        cpu: '4'
        memory: 8Gi
    requests:
        cpu: '2'
        memory: 4Gi