Remote Jupyter Session Management Tool

rjy Rust command-line tool that manages SSH tunneling for working with multiple Jupyter notebooks and lab instances over SSH. The command-line tool spawns a background SSH process and manages session information via a cache in ~/.remote_jupyter_sessions.

First, create a remote Jupyter session on a server with,

$ jupyter lab --no-browser --port=8904

where 8904 is a random high port number.

Then, copy the link it provides and use it with rjy new <link> and your remote server hostname to register this session with your local computer,

$ rjy new http://localhost:8904/lab?token=b1fc6[...]b7a40 remote
Created new session ponderosa:8906.

You could use an IP address too, but I strongly recommend if you interact with servers a lot over SSH, you add them to your ~/.ssh/config file (see this page, for example) and refer to them by their hostnames. You also should use ssh-add, so that you won't be prompted for a password each time. These may seem like frustrating extra steps, but both of these tips will greatly simplify working with remote servers a lot!

Then, we can see this Jupyter session is "registered" and the SSH tunneling is with rjy list:

$ rjy list
 Key (host:port) | Process ID | Status    | Link                              
-----------------+------------+-----------+-----------------------------------------------
 ponderosa:8906  | 68190      | connected | http://localhost:8906/lab?token=5e2f[...]8467
 sesame:8906     | 67087      | connected | http://127.0.0.1:8906/lab?token=3aa1[...]bee1

Most good terminals will allow you to directly click this link (e.g. in iTerm2 on Mac, if you hold ⌘ and hover over a link, it will become clickable).

We can disconnect a session with rjy dc <key>, where the key is that in the list output. If no key is specified, all sessions are disconnected.

$ rjy dc remote:8904
Disconnected 'sesame:8906' (Process ID=67087).

Now we can see it's disconnected:

$ rjy list
 Key (host:port) | Process ID | Status       | Link                              
-----------------+------------+--------------+-----------------------------------------------
 ponderosa:8906  | 68190      | connected    | http://localhost:8906/lab?token=5e2f[...]8467
 sesame:8906     |            | disconnected | http://127.0.0.1:8906/lab?token=3aa1[...]bee1

We can reconnect with rjy rc. Without a key, everything registered is reconnected. With a key, only that session is.

$ rjy rc remote:8904
Reconnected session sesame:8906.

Now if we check,

$ rjy list
 Key (host:port) | Process ID | Status    | Link
-----------------+------------+-----------+----------------------------------------------------------------------------------
 sesame:8906     | 69233      | connected | http://127.0.0.1:8906/lab?token=3aa1[...]bee1
 ponderosa:8906  | 68883      | connected | http://localhost:8906/lab?token=5e2f[...]8467

it's reconnected as expected. Finally, to drop a session from the registered cache (kept in ~/.remote_jupyter_sessions), use rjy drop <key>:

$ rjy drop ponderosa:8906
Disconnected 'ponderosa:8906' (Process ID=68883).

You can also drop all connections with rjy drop --all. See the built-in instructions with rjy --help for more information.

Security

This stores the token Jupyter creates in ~/.remote_jupyter_sessions, and sets the permissions so only the owner has read/write permissions. This is as secure as having the authentication token in your shell history, but caution is still warranted. Do not use on untrusted systems.

Install

$ cargo install remote_jupyter