Open talajasi7 opened 3 years ago
Thanks @talajasi7 this is a great idea, and we will definitely add it.
For future reference, currently you can stop a specific instance of the agent if you use the exact same command and add --stop
at the end.
Example, launching two agents (one per gpu), then stopping agent on gpu "1"
clearml-agent daemon --queue default --gpus 0
clearml-agent daemon --queue default --gpus 1
clearml-agent daemon --queue default --gpus 1 --stop
Thanks @talajasi7 this is a great idea, and we will definitely add it. For future reference, currently you can stop a specific instance of the agent if you use the exact same command and add
--stop
at the end. Example, launching two agents (one per gpu), then stopping agent on gpu "1"clearml-agent daemon --queue default --gpus 0 clearml-agent daemon --queue default --gpus 1 clearml-agent daemon --queue default --gpus 1 --stop
Such a command consistently stops the wrong agents from wrong queues.
@ColdTeapot273K Perhaps best to open it as an issue for clearml-agent?
@ColdTeapot273K Which agent version are you using? I just tried the same sequence of commands with the latest agent version (v1.7.0) and it consistently stops the correct agent
For now, the only way I know to kill a particular agent from ClearML API is through the arguments that define it (those that are entered when spinning it up). In addition to this feature, I think the preferable way to stop an agent would be via command-line option like:
clearml-agent daemon --stop AGENT_PID
It would make a good combination with the list of agent IDs provided by theclearml-agent list
command.