Skip to main content

gnomAD Project Application Deployment

gnomAD Browser​

System Requirements​

The gnomAD browser contains several components. The requirements for each are listed independently below.

  • Elasticsearch: 4 data nodes, each with
    • vCPU: 8
    • RAM: 64GiB
    • Disk: 1.7TB SSD (GCP Documentation)
      • Total cluster storage: 6.8TB
  • Redis:
    • vCPU: 1
    • RAM: 48GiB
    • DiskP 100GB SSD (GCP Documentation)
    • Note: This is the fewest number of CPUs google will provide with this amount of RAM
  • API:
    • vCPU: 2
    • RAM: 12GiB
  • Browser (frontend HTTP):
    • vCPU: 0.5
    • RAM: 512GiB
    • Disk: 20GiB SSD (for nginx HTTP cache)

Development / Demos​

Production​

Exome Results Browsers​

The exome results browsers are deployed on GKE, with manifests an environments managed by Kustomize. Configurations can be found in the [gnomad-deployments](https://github.com/broadinstitute/gnomad-deployments repository).

System Requirements​

Note, CPU and RAM are upper bounds. In practice, we've found that the app uses less on average.

Development / Demos​

To create a new demo environment, simply make a copy of the 'demo' kustomization, and edit it to meet your needs. That demo deployment example includes a few extras, which you may find useful to adapt:

  • an Ingress, to expose your demo on a public IP address
  • A patch to change the name of a new persistent data disk (you may need this if you are testing updated or new datasets.
$ cp -r demo my-demo
# make edits as needed
$ vim my-demo/kustomization.yaml
$ vim my-demo/ingress.yaml

# view your manifests
$ cd my-demo
$ kustomize build .
# deploy to a kubernetes cluster
$ kustomize build . | kubectl apply -f -
# or, in the case of an already existing deployment,
# diff your local changes with the active deploy:
$ kustomize build . | kubectl diff -f -

Production​

The production deployment of the exome results browsers can be found in the gnomad-deployments repository, under exome-results-browsers/prod.

In the most common case of a simple image update, the deployment manifest is updated automatically:

  1. After merging a pull request, a docker image is built via Google Cloud Build.
  2. A new image with the exome-results-browser's git commit (short) SHA is built and pushed.
  3. After a successful image build, the prod deployment kustomization is updated with the new tag in the gnomad-deployments repository.
  4. When you're ready to deploy, navigate to the exome-results-prod app in Argo, and click "Sync".