The main goal of Kuber is to help with massively parallel computations. It leverages kubernetes and docker in order to create a container that automatically runs orchestrated tasks in parallel via expansion. If you already use Google Cloud Platform, Kuber is also able to automatically create clusters, run computations, and track their progress with Google’s cloud SDK.

Before installing

To take full advantage of Kuber, you must use GCP. Creating an account is easy and you also get a bunch of money to use in your 1-year trial, but note that we are not responsible for any expenses you incur as a result of using Kuber! This is FOSS, so be very careful.

The first step should be creating an account. Just head over to the free trial page and register a Google account. In just a few seconds Google will create a new project (with a randomly generated ID) that can be explored through the console.

There are loads of different solutions available, and, before using any of them, reading their documentations might be helpful. To understand how Kuber works, there are three main pages to look through: Google Compute Engine (GCE), Google Kubernetes Engine (GKE), and Google Cloud Storage (GCS).


The installation process is, at the moment, very bare bones. Installing the package itself is simple because it has no mandatory dependencies, but installing the command line tools it leverages can be a bit confusing. The package works by issuing Unix commands, so it is probably not going to work on Windows.

If you’re using Debian/Ubuntu, Kuber has the very convenient kub_install_docker() and kub_install_gcloud() functions that should setup the environment just right. But for any other platforms you’ll have to setup everything yourself.

Strictly speaking, Kuber needs to run three commands without administrative privileges: docker, gcloud and kubectl. Documentation on how to install them can be found at their respective websites:

Sudo and docker

At least on Linux, Docker commands require sudo privileges. In order to solve this minor issue, please refer to their documentation. Remember that Kuber is not going to work unless it can execute without sudo.

Gcloud init

In order to use the gcloud command, you must first allow your account to access some APIs. In the API dashboard, click on “Activate APIs and services” and choose “Compute Engine API”, “Kubernetes Engine API” and “Cloud Storage API”.

After installing Google’s Cloud SDK and setting up API access, make sure you run gcloud init on your terminal afterwards in order to authenticate your computer as a trustworthy device on GCP. Choose any default zone you want, but make sure to choose one.

Service account

Unless your parallel task doesn’t return any results, you’ll want to save its output to a persistent storage medium (which GKE doesn’t provide in a straightforward way). After installing and making sure everything is up and running, the only step left is creating a service account in GCP so that any Docker image can write to a persistent GCS bucket.

This process is very simple and can be done in the IAM tab. Give the service account a memorable name (like “storage”), leave the ID as is and describe your Kuber needs. Follow on to the next screen and chose the account’s roles; they can be as broad as you want, but it must include “Storage Admin”, “Storage Transfer Admin”, and “Storage Object Admin”. In the last screen just click “Create Key”, select JSON as the type and accept the changes.

The result of this process will be a file known as a “service account key” which can be used in any gcloud command as an authentication on your behalf. Be very careful with this file because it can be used to create/delete anything in your GCS account.

All set

Now Kuber’s setup process is complete. To learn more about how to create a parallel task, take a look at the “Toy example” vignette.