Managing HTC-Grid after deployment

Managing the HTC-Grid agent

1. Make connection with EKS, go in the deployment/grid/terraform folder where terraform apply has been run. Then make sure that KUBECONFIG is set (see step 6).

2. Set up the storage for helm

   export HELM_DRIVER=configmap
   

3. Deploy the agent

   helm install <your release name> ../charts/agent-htc-lambda --set fullnameOverride=htc-agent
   

4. Test the installation by running

   helm test <your release name>
   

5. Delete the agent

   helm uninstall <your release name>
   

6. Get deployed pods ( see notes after agent deployment )

7. Get the log of a pod: ( see notes after agent deployment )
8. Describe the state of a pod , useful for debugging situation never run, (see notes after agent deployment)

9. Execute a command into a pod

   kubectl exec  <pod name> -c <container name> <command>
   

10. Open an interactive session into a pod

    kubectl exec -it <pod name> bash
    

11. Get config-map used in the pods, i.e., current configuration

    kubectl get cm agent-configmap -o yaml
    

12. Get information about the pod autoscaler

    kubectl get hpa
    kubectl get hpa -w
    

13. Updating number of replicas (my_cluster is your release name that you selected with helm)

helm upgrade --set replicaCount=5 --set foo=newbar my_cluster ./agent-htc-lambda

Common commands

• Get logs from a running agent and lambda worker

  kubectl logs -c <agent or lambda> <pod name>
  
  Examples:
  kubectl logs -c agent htc-agent-544bd95456-wgzqs
  
  kubectl logs -c lambda htc-agent-544bd95456-wgzqs
  

• Launching Tests

  cd root/examples/submissions/k8s_jobs
  kubectl apply -f <test-name.yaml>
  
  example:
  kubectl apply -f scaling-test.yaml
  
  Follow the execution of the test
  kubectl logs job/scaling-test -f
  

• Deleting all running pods

  kubectl delete $(kubectl get po -o name)