I am using two different environment for mlops and model training. I have setup the agent in the training environment. I can start the queue using command 'clearml-agent daemon --queue default' and that works perefectly, whenever I run a pipeline from mlops it works fine.
But now I am trying to implement a startup script for 'clearml-agent daemon --queue default' so that whenever the training system starts/restarts the agent queue also starts. I used /etc/rc.local and added a single line to execute a single line script 'clearml-agent daemon --queue default'. But in this case queue starts but does not work properly, whenever I am using any nvidia tao command through pipeline I get a error "The input device is not a TTY".
However if I run clearml-agent daemon --queue default from terminal; then all tao commands from clearml pipeline works perfectly. DOnt understand why it does not work when setup as startup script using rc.local.