Skip to content

Deployment of WorkFlow Launcher⚓︎

Make a deployment⚓︎

The WorkFlow Launcher is currently running on a Kubernetes cluster, and the deployment is managed by Helm. In order to make a deployment or upgrade the current deployment manually without using the automations, you have to make the following preparations:

Setup Kubernetes⚓︎

  • Make sure you have the kubectl command available, while in the Broad network or using the VPN, run a command like the following to set up the connection to the desired cluster:
gcloud container clusters get-credentials \
  gotc-dev-shared-us-central1-a --zone us-central1-a \
  --project broad-gotc-dev

  • Run kubectl config get-contexts to make sure you are connected to the right cluster.

  • Later you could run kubectl config use-context $CONTEXT_NAME to switch (back) to other contexts.

Setup Helm⚓︎

  • Install Helm, please follow the Helm instructions or simply try brew install helm, assuming you have HomeBrew installed on your MacOS.

  • Run:

helm repo add gotc-charts https://broadinstitute.github.io/gotc-helm-repo/
to add gotc’s Helm repo to your Helm. Note gotc-charts is just an alias, you could give it any name you want.

  • Run:

helm repo update
to make the local cached charts update-to-date to the remote repo and also run:

helm repo list
to check the list of repo you have connected to anytime you want.

  • In the Broad network or on VPN and your kubectl is setup to connect to the right cluster, run:

helm list
to check the current deployments that are managed by Helm.

Build and Dockerize⚓︎

Build the WFL with make as it's described in Quickstart:

$ make ui api TARGET=images -j8

Info

To manually compose a new docker image if you want one. Tag it with some helpful name, then push it to DockerHub so Kubernetes can find it, run:

docker build -t broadinstitute/workflow-launcher-api:$IMAGE .
docker build -t broadinstitute/workflow-launcher-ui:$IMAGE  ui/.
docker push broadinstitute/workflow-launcher-api:$IMAGE
docker push broadinstitute/workflow-launcher-ui:$IMAGE

You should see it listed here:

https://hub.docker.com/repository/docker/broadinstitute/workflow-launcher-api

Clone the gotc-deploy repository.⚓︎

git clone --branch <the branch you want> \
  https://github.com/broadinstitute/gotc-deploy.git

Render the chart⚓︎

Render the wfl-values.yaml file.

docker run -i --rm -v "$(pwd)":/working \
  -v "$HOME"/.vault-token:/root/.vault-token \
  -e WFL_VERSION=latest \
  broadinstitute/dsde-toolbox:dev \
  /usr/local/bin/render-ctmpls.sh \
  -k ./gotc-deploy/deploy/gotc-dev/helm/wfl-values.yaml.ctmpl

Note: That command always fails, so look at ./gotc-deploy/deploy/gotc-dev/helm/wfl-values.yaml to verify that the values substituted into the template are correct.

Note: Some of the values in these YML files contain credentials or sensitive information, so DO NOT check them in to your version control system or make them public!!!

Deploy⚓︎

Run this to upgrade a deployment release.

helm upgrade wfl-k8s gotc-charts/wfl --install \
  -f ./gotc-deploy/deploy/gotc-dev/helm/wfl-values.yaml

That doesn't recreate ingress resources and so on, but it will --install everything necessary if a release is not deployed already.

Run kubectl get pods to review your deployments.

# kubectl get pods
NAME                                              READY STATUS  AGE
cromwell-auth-gotc-dev-authproxy-6f76598bb4-snpp6 1/1   Running 11d
hjf-test-authproxy-78758bbb6b-7lnfg               2/2   Running 42d
wfl-k8s-6948f6566d-2bdcp                          2/2   Running 2d5h
#

Run kubectl logs wfl-k8s-6948f6566d-2bdcp workflow-launcher-api to view the logs for the WFL-API server.

Run kubectl logs wfl-k8s-6948f6566d-2bdcp workflow-launcher-ui to view the logs for the WFL-UI front end.

Testing the deployment locally⚓︎

Similar to the above process, but in addition to the above preparations, you also have to:

  • Make sure you have a relatively new version of Docker client installed, which has Docker -> Preferences -> Kubernetes -> Enable Kubernetes. Turn on that options and restart your Docker client.

  • Run kubectl config use-context $CONTEXT_NAME to the Docker-built-in Kubernetes context, usually it should be called something like docker-for-mac.

Similar to how you have setup the Helm charts, you could run:

helm upgrade gotc-dev gotc-charts/wfl -f custom-authvals.yaml --install

To install or upgrade the deployments on your local cluster. One thing to note is that you CANNOT create an ingress resource locally, so it’s important to disable the ingress creation in your custom values YML files.