Helm charts for the FNNDSC and the ChRIS Project.

Chart Name	License	Chart Version	App Version	Description
fnndsc/chris	MIT			Open-source platform for medical compute.
fnndsc/pfcon	MIT			Standalone remote compute resource service for ChRIS backend.
fnndsc/orthanc	GPLv3+			Open-source PACS server. https://www.orthanc-server.com/
fnndsc/ohif	MIT			Web DICOM viewer. https://ohif.org/
fnndsc/linkerd-nodeport-workaround	MIT			Workaround for using Linkerd with NodePort services.

If you already have Docker installed, the easiest way to obtain k8s is KinD. KinD installation instructions are here: https://kind.sigs.k8s.io/docs/user/quick-start/

Development scripts are defined in testing/justfile, which uses the just syntax. You should install just: https://github.com/casey/just#installation

Then you can run things like:

git switch dev
helm dependency update ./charts/chris

cd testing

just kind
just up
just wait
just test

Optionally, to use chrisomatic:

just chrisomatic

Then, graceful tear down:

just down
just unkind

Editing pfcon's templates can be tricky because it's a dependency from the same repo as chris. Here's a workaround:

cd charts/chris/charts
rm ./pfcon-*.tgz
ln -sv ../../pfcon

To publish your changes, increase version in charts/pfcon/Chart.yaml then merge the dev branch into master. Once the release is created by GitHub Actions, increase the version and pfcon dependency version in charts/chris/Chart.yaml then update charts/chris/Chart.lock by running

cd charts/chris
rm charts/pfcon
helm dependency update .

Finally, push to master once more.

Optionally, a Kubernetes observability stack can be deployed into the Kind cluster. You can choose between OpenObserve or a Grafana-based stack.

OpenObserve is much more efficient than Grafana. I also have log collection set up for OpenObserve. However, OpenObserve's visualization and data querying is not as good as Grafana. It is recommended to use OpenObserve if you (a) need to aggregate and search through logs, and/or (b) your workstation has less than or equal to 8 CPUs, 16GB RAM.

Logs are collected using Vector and visualized using OpenObserve. To run the observability stack and open the dashboard, run

just openobserve

Log in with the email [email protected] password chris1234.

Alternatively, you can get logs from the command-line using the just olog [POD_NAME_LABEL] command. Examples:

just olog pfcon

just olog chris-heart

Two releases of Vector are made:

• "Agent" mode which runs on every node to collect logs and host metrics
• "Stateless-Aggregator" which scrapes Kubelet /metrics/cadvisor

These logs and metrics are shipped to OpenObserve.

TODO