Skip to content

Instalation TOBIAS Kubernetes S3 Version

Jump to bottom
goymann edited this page Mar 2, 2020 · 12 revisions
Clone this wiki locally

To run the Pipeline you need a S3 Account and a Kubernetesaccount. If you have, a Kuberntesaccount go to the Rancher Website of your Cluster and get the access configuration key. Store this key under '~/.kube/config'

Set up the Cluster

The setup of the cluster has been automated with a python script it creates the required persistent volume claim on the cluster and stores the S3 credentials in secrets on the Cluster. To run it you first have to install the PYKS-Package. For this go into the Package directory and run the setup script. After this go bag and start the Setup_TOBIAS_MAPOKS script.

 $ cd ../PYKS
 $ python setup.py install 
 $ cd ../TOBIAS_MAPOKS/
 $ python Setup_TOBIAS_MAPOKS.py --s3secret <S3SECRET>  --s3username <S3USERNAME>  --s3endpoint https://s3.mpi-bn.mpg.de --namespace <your namespace>

Persistent Volume Claim

The Persistent Volume Claim is a managed-nfs-storage so a nfs storage must be available on the cluster. The pvc gets name 'nextflowpvc' with a size of 250Gib if there is already a pvc called 'nextflowpvc' the script will ask you if it should replace the pvc (If you replace the pvc all your files in the current pvc get lost!). If you want to use another name for the pvc specify this with --pvc_name by running the setupskript. Then you must change also the name of the PVC in the config file of the pipeline. For more information look at link.

S3-Secret

The S3 credentials must be available on the cluster so that the Jobs are able to download and upload the files to the S3. The credentials are stored on secretes on the cluster. Secrets are objects on the cluster, which contain sensitive information’s, such as passwords and SSH keys. The default name of the secret created by the setup script id s3mpi. If you want to change it set --secretname when running the setup script and change the name of the secret in the config file of the pipeline. For more information look at link

Run the Pipeline

To start the entire pipeline, first create the Conda environment, then activate it. Then go to the configuration file and set the namespace in which the calculation should be executed (must be the same as that set for the setup script). Add your S3 credentials to the file and then activate the calculation by setting Kubernetes to true to activate the cleanup set kubernetes_clean_up to true.

#Kubernetes
kubernetes: false #when is sett true plotting runs on Kubernetes 
kubernetes_clean_up: false #when set on true all workalods and s3 buckets will delte automaticly
name_of_secret:  's3mpi'
name_of_pvc:  'nextflowpvc'
namespace: 'nextflowtest'
   #S3 Requierments
username: ''
passkey: ''
endpoint: 'https://s3.mpi-bn.mpg.de'

The cleanup Process deletes the files of the calculation in the PVC.