Skip to main content


MiniKF is Arrikto's single instance one click deployment Kubeflow offering.

Details, documentation and install instructions are availble via Kubeflow Documentation .

We hope this FAQ helps you answer MiniKF questions, we realize we don't have all the answers because we do not yet have all the questions! If you are not able to find an answer to your question below please reach out to us on the MiniKF channel on the Kubeflow Community slack.

For quick support and resolution when you reach out please share:

  1. Which instructions are you following? Please link to them.
  2. Which step is failing?
  3. What did you expect to happen?
  4. What actually happened? (logs are tremendously helpful for the community)

MiniKF Frequently Asked Questions

MiniKF Basics

➜ What is the difference between MiniKF and regular Kubeflow installation on Kubernetes/Minikube?MiniKF is a packaged version of the whole stack that installs automatically, so that you do not have to install Kubeflow manually. We believe it is the easiest and fastest way to get up-and-running with Kubeflow. MiniKF also includes Arrikto's Rok software for advanced data management, giving you data versioning, reproducible pipelines, data protection for your notebooks, and the foundation to share workloads along with data across K8s clusters.
➜ Do I need to start the VM to access MiniKF Kubeflow UI?Yes, the VM needs to be started and MiniKF needs to start up before the endpoint URL can be accessed.
➜ Do I need to run minikf in order to start MiniKF in the VM?No, you can SSH into the VM and execute minikf to see status but it will autostart.
➜ Does the login URL for the Kubeflow UI change each time VM is started?For GCP the URL will be static, for AWS it will change and it will be necessary to wait until the VM and MiniKF have started to see the new URL.
➜ I see MiniKF has been upgraded for KF 1.4, are my prior MiniKF deployments upgraded automatically?To take advantage of upgraded MiniKF you will need to redeploy via GCP or AWS Marketplace.
➜ I’m thinking of upgrading but noticed number 5 in the instructions says “Remove all local state. This will remove all of your customization in MiniKF (notebooks, pipelines, Rok snapshots)“. Is this necessary and if so, does this mean we need to manually export notebooks, datasets, pipelines, etc if we want to preserve our work?Unfortunately, we don't support MiniKF upgrades without losing the existing data. However, we do work with customers to provide enterprise Kubeflow support and/or support for their MiniKF deployments. If this describes your situation please reach out to your account manager.
➜ I am unable to start MiniKF due to account not validated error.Before launching MiniKF in either GCP or AWS you must first validate your account with the cloud provider.
➜ My docker images installed at /home/jovyan are not available?Jupyter Notebook Severs will mount a workspace volume in the home folder /home/jovyan. As a result any Docker Images stored in this directory will be overwritten. To include additional Docker Images make sure to store in a different location than the home folder.

MiniKF & Tutorials

➜ I tried to run the tutorials from and I am running into missing package errors.You must make sure the packages are imported by running the first cell or any import commands and then restarting the notebook kernel. As an example consider reviewing our Kale 101 Course - Get Dataset and Code section

MiniKF & Vagrant

➜ I am experiencing massive memory issues when running the Titanic example on my laptop (MiniKF on Vagrant).The Titanic example requires higher memory than the MiniKF VM on Vagrant has, you could upgrade your VM's RAM, but make sure that the host has enough RAM to support this. The easiest way is to run the example using MiniKF on GCP:
➜ I started MiniKF on a personal / private server and would like to share it on the network. What's the proper way of doing it? I tried forwarding port with Vagrant: "forwarded_port", guest: 80, host: 8081 # MiniKF Terminal "forwarded_port", guest: 8080, host: 8080 # KubeFlow "forwarded_port", guest: 7681, host: 7681 #
➜ Although MiniKF Terminal is available, Kubeflow UI is not responding.Add in guest_ip: "" in the above lines "forwarded_port", guest: 80, host: 8081 # MiniKF "forwarded_port", guest: 8080, guest_ip: "", host:8080 # kubectl "forwarded_port", guest: 7681, host: 7681 #
➜ I cannot install MiniKF on my laptop.Your system must been the following requirements.
➜ I have a MacBook that I believe meets the requirements, but I cannot install MiniKF.While we test MiniKF on various MacBook models we can't test it exhaustively on all of them. With that said in general consider:
  1. Monitoring your CPU utilization and check if your Macbook overheats when starting MiniKF.
  2. Consider using a faster machine, if you have one available.
  3. Try MiniKF on GCP.
➜ Can I expose the MiniKF UI to be accessed from outside the host?You will need to forward the necessary ports to the host, and possibly create some firewall rules. See the relevant Vagrant docs:
➜ On my machine, I have 128G memory and 28 CPUs. By default, the Minikube installation of the Vagrant Box uses 2 CPUs and 10G of memory. How can I increase this to 8 CPUs and 32G of memory. I amended the VM specs from within the VirtualBox UI.Although you can edit your VM from within the VirtualBox UI, you shouldn't, because Vagrant will overwrite [at least some of] your changes based on what is in the Vagrantfile. So, it's best to edit the Vagrantfile and specify the CPU and RAM you need.
➜ During install Vagrant hangs at pod 33 / 34.This is likely due to a memory issue, make sure that the Vagrantfile is providing 24 GB (okay) - 32 (best) GB of Ram for Kubelow install.
➜ How do I access MiniKF behind a proxy?To access MiniKF behind a proxy you can use the vagrant-proxyconf plugin and set Vagrantfile accordingly after init. This linkprovides more info regarding proxy configuration for Vagrant.
➜ I cannot install MiniKF. I get the following error message: VT-x is not available (VERR_VMX_NO_VMX)
Hyper-V disables VT-x for other hypervisors, so VirtualBox cannot use it. Hyper-V disables VT-x for other hypervisors and VirtualBox says there is not VT-x available because when hyper-V is installed on Windows, the hypervisor is running all the time underneath the host OS. Only one process can control the VT hardware at a time for stability. The hypervisor blocks all other calls to the VT hardware. To proceed, you need to disable Hyper-V:
  1. Open CMD as administrator.
  2. Turn off Hyper-V by running: /set hypervisorlaunchtype off.
  3. Reboot.
To turn it back on:
  1. Run bcdedit /set hypervisorlaunchtype.
  2. Reboot.
➜ After completing Vagrant install unable to open due to timeout.Destroy and restart the VM to resolve the issue.
➜ While following the Vagrant install steps here Vagrant returns an error The IP address configured for the host-only network is not within the allowed ranges.
This is a Vagrant issue and results from using a version after 6.1.26 ( To correct this please use version 6.1.26 and make sure you are following the most recent directions:
➜ Can I use GPUs in MiniKF in Vagrant?Yes, details on this topic can be found here:

MiniKF & GCP

➜ When I try to get MiniKF to run on GCP, I get resource level errors.You cannot run MiniKF (or Kubeflow in general) using the GCP Free Tier. See the official docs:
➜ Can I use GPUs in MiniKF in GCP?Yes, please note you normally requisition a VM with one GPU but don't assign that GPU to your notebook server. You specify if GPUs are necessary when tagging cells with Kale.
➜ When I try to get MiniKF to run on GCP, I get an QUOTA_EXCEEDED error.For default values, MiniKF requires 200G for boot disk and 500G for data disk. You will need to select a region with a greater quota limit. Please request a quota increase for your account if none are available.

MiniKF & AWS

➜ Can I use GPUs in MiniKF in AWS?Yes, please note you normally requisition a VM with one GPU but don't assign that GPU to your notebook server. You specify if GPUs are necessary when tagging cells with Kale.

MiniKF & Windows

➜ If I have to install Kubeflow on Windows should I install MiniKF only or is there is something else I can install?You should only install MiniKF assuming you already have Vagrant and VirtualBox on your windows machine.

MiniKF & MacOS

➜ I’d like to try MiniKF on macos, but I don’t want to install VirtualBox. How can I get it working with HyperKit or Docker?VirtualBox is a prerequisite for using MiniKF on your laptop. As an alternative, you can use MiniKF on GCP:

MiniKF & Notebooks

➜ When I do a PIP install, it says permission denied. When I open a terminal through Jupyter notebook and run sudo pip install, it asks for a password for the account jovyan.Please run this command from inside a Notebook Server:
pip3 install --user <library>

MiniKF Logs

Collecting MiniKF logs depends on where MiniKF is being used.


  1. vagrant ssh into your MiniKF VM.
  2. run minikf-gather-logs. This will produce a tarball .tgz file in your MiniKF directory.
  3. Open the tarball file to see logs.
  4. When necessary and instructed share the tarball file with Arrikto experts.


For GCP, in general, to collect and review logs you must:

  1. SSH from the Browser window.
  2. run minikf-gather-logs. This will produce a tarball .tar.bz2 file under /vagrant/.
  3. Download the file as described in this guide.. The exact path to download the logs file is /vagrant/minikf-logs-<date>-<time>.tar.bz2.
  4. Open the tarball file to see logs.
  5. When necessary and instructed share the tarball file with Arrikto experts.