@<1523701087100473344:profile|SuccessfulKoala55> Thanks for the clarification; it helped! Is there a way to run directly on the k8s agent using the CLI or another method without cloning an existing task and enqueuing it?

Hey @<1523701205467926528:profile|AgitatedDove14> ,
I'm using aws_autoscaler.py to create new EC2 instances.
Yup, it does the job of spinning up EC2 instances. You can verify it from the logs where it shows the EC2 instance with ID i-0b4ee42a4fafd220d being created.

![image](https://clearml...

solution can be found here: None

added these to the agent deployment but still not working

@<1523701087100473344:profile|SuccessfulKoala55> logs attached, i use agent version 1.24-21, from the installed packages i see we use clearml==1.12.1

@<1523701087100473344:profile|SuccessfulKoala55> never mind I found the answer to use task.execute_remotely .. ty

Hi @<1523701070390366208:profile|CostlyOstrich36> it fails i got Process failed, exit code 1

solved using adding this to the agent config:

clearmlConfig: |-
    agent{
       extra_docker_shell_script: ["apt-get install -y graphviz", ]
    }

config file: None

@<1523701087100473344:profile|SuccessfulKoala55> could you please explain how to mount it in agent conf, do you have an example..

@<1523701070390366208:profile|CostlyOstrich36> Thanks for responding. In the video, the explanation about running the experiment in the Kubernetes agent was very brief, and no demo was provided for that.
Here are the steps I followed considering that I'm connected to the cluster:

• I installed the ClearML agent Helm chart, and now I have the ClearML agent running in the "clearml" namespace.
• After that, I ran the script from the following GitHub repository: [None](https://github.com/allegr...