How to use and install my own local Phenomenal Galaxy infrastructure ?
For the development of a tool in PhenoMeNal, for curiosity or for fun, this gist explains step by step how to get a local up and running PhenoMeNal e-infrastructure on Kubernetes minikube.
Check out repo
git clone https://github.com/phnmnl/container-galaxy-k8s-runtime.git
cd container-galaxy-k8s-runtime
Start minikube
minikube start
Configure helm
helm init
helm repo add galaxy-helm-repo https://pcm32.github.io/galaxy-helm-charts
Reconfigure docker to use minikube docker deamon and install your own galaxy image
eval $(minikube docker-env)
Get a terminal
Stand on the directory where the galaxy dockerfile is
Make sure that you're in the correct branch and commit
Then doeval> $(minikube docker-env)to use the docker daemon of minikube
Build your container
docker build -t galaxy-test .
Create an admin role to enable launching jobs
kubectl create clusterrolebinding add-on-cluster-admin --clusterrole=cluster-admin --serviceaccount=default:default
Install and run custom helm
helm install --set galaxy_image="galaxy-test", galaxy_image_registry="", development_folder="/hosthome/path/to/container-galaxy-k8s-run", pv_minikube="yes", galaxy_admin_email="[email protected]", galaxy_admin_password="admin1", galaxy_api_key="qwertyuio" galaxy-helm-repo/galaxy
"galaxy-image" can be customized.
"development_folder" needs a path to your local galaxy repo. On linux systems "home" directory is replaced by "hosthome" in the virtual environment, so paths should take it into account.
"galaxy_admin_password" requires at least 6 characters.
Check the running pod
kubectl get pods
kubectl logs -f <galaxy-k8s-NUMBER from previous command>
When everything is set up and running, the log should give:
...
serving on 0.0.0.0:8080 view at http://127.0.0.1:8080
You can then open the Galaxy instance in your browser with:
open http://$(minikube ip):30700
Clean up (actually wipe out) minikube, helm and everything, and restart everything nicely from scratch (execute each command one by one)
minikube stop
minikube delete
rm -r $HOME/.helm
rm -r $HOME/.minikube
minikube start
minikube ssh
sudo rm -rf /data/galaxy-data
exit
helm init
helm repo add galaxy-helm-repo https://pcm32.github.io/galaxy-helm-charts
eval $(minikube docker-env)
docker build -t galaxy-test .
kubectl create clusterrolebinding add-on-cluster-admin --clusterrole=cluster-admin --serviceaccount=default:default
helm install --set galaxy_image="galaxy-test", galaxy_image_registry="", development_folder="/hosthome/path/to/container-galaxy-k8s-run", pv_minikube="yes", galaxy_admin_email="[email protected]", galaxy_admin_password="admin1", galaxy_api_key="qwertyuio" galaxy-helm-repo/galaxy
If you want to restart your Galaxy runtime container, do somethings like
helm list
helm delete <funny-name>
kubectl delete secrets/galaxy-admin-secret
kubectl delete configmaps/galaxy-admin-user
kubectl delete configmaps/db-connection #if necessary
kubectl delete secrets/galaxy-postgres-secret #if necessary
docker build -t galaxy-test .
helm install --set galaxy_image="galaxy-test", galaxy_image_registry="", development_folder="/hosthome/path/to/container-galaxy-k8s-run", pv_minikube="yes", galaxy_admin_email="[email protected]", galaxy_admin_password="admin1", galaxy_api_key="qwertyuio" galaxy-helm-repo/galaxy
Restart galaxy interface directly from inside the container
con=$(kubectl get pods | grep Running | cut -d" " -f1)
kubectl delete pods $con
sleep 4
con=$(kubectl get pods | grep Running | cut -d" " -f1)
kubectl exec -i -t $con -- /galaxy/run.sh --restart
Delete all non running or container creating pods
kubectl get pods | grep -v -E "ContainerCreating|Running|NAME" | awk '{print $1}' | xargs kubectl delete pod
Delete all jobs done (successful = 1)
kubectl get jobs | awk '$3 !~ 0 && $1 ~ /^galaxy*/ {print $1}' | xargs kubectl delete jobs
See ports listening
netstat -tulpn
or
ps -ef
See status of the docker service
systemctl status docker.service
# to start / stop the docker daemon
service docker start
service docker stop
After helm init, you should be able to run
kubectl get pods --namespace kube-system
and see Tiller running.