(TA) Adding Access from Pi to Training Node

Since we cannot train the model on the pi, we have to send all of the data we collect to the instance we are using for training.

Note: commands from steps 1-3 are run inside the raspberry pi container, steps 4-5 are run on the baremetal instance used for training, and steps 6-10 are run back on the pi.

Syncing the Data:

1. In order to allow the pi to connect to the chameleon node, we must give it a key pair to enable an ssh connection. Run the following command in the pi container.

ssh-keygen -t ed25519

Press enter to whatever prompts come up so that the key is saved in the ~/.ssh folder and there is no passphrase needed to use it.

1. Change the permissions on the private key so that it can be used with ssh

chmod 600 ~/.ssh/id_ed25519

1. View the public key that was generated and copy the entire key.

cat ~/.ssh/id_ed25519.pub

Note: be very careful when selecting the key with your cursor. You want to make sure you are not copying any additional whitespace or leaving out any characters.

1. Back in the terminal of the training node (you may need to ssh cc@<ip address> into it again from your own terminal), open the authorized_keys file.

nano ~/.ssh/authorized_keys

1. Paste the key you copied at the bottom of this file.

   1. Make sure that when the key is pasted in, it is all on one line and that there are no missing characters or extraneous whitespace

2. Back on the pi, test the ssh connection by running the ssh command to connect to the ip address of the training node. This is the same command you would run on your own computer to connect to the training node.

ssh cc@<ip address of training node>

1. If the last step was successful, type "exit" to leave the ssh connection and return to the pi.