TODO: @Xing, Marjorie Lucas @alphan, Alexander Torres

Developer Considerations

1. The node is still subject to spot terminations, so don't depend on it staying up forever

2. The developer is now responsible for the life cycle of this node, to turn it off when it's not use and to terminate it when it's no longer needed

3. The developer should probably relabel the node, if possible, so other developers know who is responsible for that instance, if any issues crop up.

4. What is the impact on the operator, who may see this node in metrics and figaro?

Key Steps

(1) detach from ASG

1. Go to the AWS EC2 console

2. Click on the Auto Scaling Groups link on the left hand side column (usually towards the bottom)

3. Click on the autoscaling group that your Verdi worker instance is in.

   1. After you click on the autoscaling group, a menu should show up on the bottom of the screen.

4. Click on the Instance Management tab and locate your Verdi worker under Instances.

5. Right click the instance you want to detach and click detach.

(2) disable harikiri

1. run ps -x in the verdi worker instance and look for the PID that contains harikiri

2. kill the process with the following command: kill <harikiri_pid>Create a file with touch /data/work/.harikiri. Harikiri will not terminate the instance as long as such a file exists.

(3) docker run bash

1. ssh into worker node

2. cd into the work directory

3. Edit the _run.sh script

   1. Edit the block of code starting with docker run ...

   2. append -it just after “docker run”

   3. remove PGE command (/home/ops/...) at the end of the docker run ... block and add /bin/bash

4. Save _run.sh changes

5. Run _run.sh