Skip to content

bromberglab/bio-node-docs

Repository files navigation

Bio-Node

Documentation for Bio-No.de

What is Bio-Node?

Bio-Node is a platform to run all sorts of bioinformatics workflows in a reliable and reproducible way. Working with kubernetes, the underlying compute cluster is rescaled automatically to account for changing performance needs. Optimizing speed and cost simultaneously.

All steps in a Bio-Node workflow are abstracted as Docker containers. This allows for easy deployment and quick maintenance & replacement. Bio-Node aims at helping computational biologists to leverage powerful computing resources, without extensive HPC training requirements. Assembling a workflow is as easy as dragging the sub-components around and configuring all settings with the visual interface. Docker also enables the system to use a lot of pre-existing tools without additional deployment efforts. All tools already existing as Docker images can be run on Bio-Node, as soon as in- and outputs of the image are set.

How do I install my own deployment of Bio-Node?

::: tip With every link to the GCP console, make sure you're navigated to the right project and account before changing anything.

Running GKE will cost you money. :::

The install script addresses this task. To create a custom deployment, follow these steps:

  1. A GCP project is needed for Bio-Node. Bio-Node uses GKE for easy kubernetes deployment.
    1. If you don't have one set up yet, create a new project.
    2. (GCP projects take a while to initialize. Make sure the project is fully created before continuing.)
  2. Install the requirements.
    1. Running the install script will check all requirements and tell you which ones are missing.
    2. gcloud sdk is needed to interact with your GCP resources. Make sure you set your config to the desired project with
      gcloud config set project [project-id]
  3. Since we work with GKE, the API has to be activated beforehand.
    1. The gcloud command to create a cluster will fail if the API was not activated.
    2. In this case, a URL will be printed where it can be turned on.
    3. As an alternative, just visit the webpanel before running the install script.
    4. Once the create cluster button is no longer grayed out, you can continue with the installation.
  4. As of the time of writing, GCE requires a service account in your project to function. Check the console to see if an account exists with the name
    [project-id][email protected]
    And has editor permissions.
    1. If this account is missing, your project is in an inconsistent state. Check with GCP support.
  5. Run the install script (after setting execution permissions with chmod +x [file])
  6. Validate all settings and validate that the account that's currently authenticated with gcloud is also associated with your project.
    1. To change the account, use gcloud auth login
    2. To change any of the settings, edit .bio-node.config
  7. (Optional) If you have a domain which you can point to the public IP of the cluster, use the Ingress access mode.
    1. Changing DNS settings takes hours, and there is multiple things than can go wrong. It's safer to set this up separately using the instructions from Google.
    2. An ingress is the easiest way to achieve HTTPS encrypted traffic reliably.
  8. (Optional) Sendgrid can be used as an email server for communications like password reset emails.
  9. After the setup is complete, make sure to navigate to the URL /api/createadmin
    This will show the output
    admin:<password>
    will tell you how to login to the admin panel with superuser privileges.

::: danger This URL is one time use only!

After the first visit to /api/createadmin the password is not retrievable without re-installing Bio-Node. :::