Running byok8s with minikube¶

Installing¶

See the Installing page for details about installing byok8s and its prerequisites (including minikube).

We cover two scenarios:

bare metal machine, i.e., a laptop or desktop machine that can run a hypervisor like VirtualBox
cloud machine, i.e., AWS EC2 node, which is itself a virtual machine and cannot run a hypervisor

These quickstarts assume you have Python and minikube installed, and that you have cloned and installed byok8s at ~/2019-snakemake-byok8s/.

Quickstart on Bare Metal Machine¶

On a bare metal machine, the procedure is relatively uncomplicated: we create a cluster, we export some variables, we run the workflow, we tear down the cluster:

# Start a minikube cluster
minikube start

# Verify k8s is running
minikube status

# Export AWS credentials
export AWS_ACCESS_KEY_ID="XXXXX"
export AWS_SECRET_ACCESS_KEY="XXXXX"

# Run the workflow
byok8s workflow-alpha params-blue --s3-bucket=mah-bukkit 

# Stop the minikube cluster
minikube stop

Quickstart on Cloud Machine¶

As mentioned above, cloud compute nodes are virtual machines themselves and cannot run a hypervisor, so things are a bit more complicated.

To tell minikube not to use a virtual machine driver, run the following command in a terminal to create a minikube config file:

cat <<'EOF' > ~/.minikube/config/config.json
{
    "vm-driver": "none"
}
EOF

Now you can start up a minikube cluster.

There is an additional DNS problem that needs to be fixed in the containers before you proceed. You will know there is a problem if you run the get pods command with kubectl and see your CoreDNS containers in a CrashLoopBackOff state:

$ kubectl get pods --namespace=kube-system
NAME                               READY   STATUS             RESTARTS   AGE
coredns-86c58d9df4-lvq8b           0/1     CrashLoopBackOff   5          5m17s
coredns-86c58d9df4-pr52t           0/1     CrashLoopBackOff   5          5m17s
...                                ...     ...                ...        ...

To fix the problem with the DNS settings, we have to patch the CoreDNS image being used by kube-system. To do that, use the file test/fixcoredns.yml test/fixcoredns.yml in this repository with kubectl apply:

# Fix the DNS container
kubectl apply -f fixcoredns.yml

# Delete all kube-system containers 
kubectl delete --all pods --namespace kube-system

The kube-system containers will be re-spawned by the cluster control system. It should happen in a few seconds, and then you'll be ready to run byok8s:

# Return to our virtual environment
cd ~/2019-snakemake-byok8s/test/
source vp/bin/activate

# Verify k8s is running
minikube status

# Export AWS keys for Snakemake
export AWS_ACCESS_KEY_ID="XXXXX"
export AWS_SECRET_ACCESS_KEY="XXXXX"

# Run byok8s
byok8s workflow-alpha params-blue --s3-bucket=mah-bukkit

Congratulations! You've just run an executable Snakemake workflow on a minikube kubernetes cluster.