2
  • I have a Windows machine where I start an IPython kernel (to do some stuff not possible on another machine).
  • I have a Linux machine from which I would like to connect to the IPython kernel running on the Windows machine.
  • I can SSH from the Linux machine to the Windows machine (using this solution: https://superuser.com/a/172299).

I have tried following: https://github.com/ipython/ipython/wiki/Cookbook:-Connecting-to-a-remote-kernel-via-ssh. Both the automatic and the manual solution gives the following:

"ERROR: Kernel did not respond"

Using the same solution, I can connect from my Linux machine to an IPython kernel running on a Linux server. Any solution to get this to work with Linux to Windows?

Community
  • 1
  • 1
Anders Lundstedt
  • 131
  • 9

2 Answers2

2

You don’t need SSH to connect to a remote ipython kernel, regardless of whether it’s a ipython kernel running on Windows or Linux or Mac. What you do need is to have the Ip of the remote kernel visible to the terminal from which you are trying to connect from. Here are the steps:

  1. Find out the ip address of the server (the machine on which the ipython kernel is running i.e. where you want the computation to happen) and client (the machine from which you are trying to connect to):

    1.1. If you are on Windows, open up the command prompt and do a ipconfig to find out the ip addresses. If the Windows server has a direct Internet connection/lan connection, you should see a couple of ips like 192.168.57.1 and 10.2.3.64 and 127.0.0.1.

    1.2. If you are on linux, open up a terminal and type ifconfig or ip addr show. You should again see a couple of ips like 192.168.57.1 and 10.2.3.64 and 127.0.0.1.

    1.3. Test that atleast one of your server ip addresses is visible from the client: Ping your server from your client, using the command ping. ping will work on either Windows or Linux terminals. If you are running the windows/Linux as a VM or is behind a firewall, it is very much possible that your client or server is not visible from the other side. You don’t have to ping the ip address 127.0.0.1. This is a loop back address, and is only visible from the same machine where you got this ip address from. For example if you ping 127.0.0.1 from the Windows machine, it will ping the same Windows machine. If your client and server instances are running on the same machine, then its fine to use this address. However, if your client or server is running on a VM or a different machine altogher, then 127.0.0.1 wont work.

  2. Start the remote kernel:

    2.1. Once you have figured out which ip address on the server is visible from the client, start a kernel on the machine using ipython kernel. The ipython kernel will startup and show that `To connect another client to this kernel, use: --existing kernel-1234.json

    2.2. Locate the kernel-1234.json file on your server by importing (https://stackoverflow.com/a/35094772/4752883) In [1]: from jupyter_client import find_connection_file In [2]: find_connection_file() Out[2]: 'C:\\Users\\me\\AppData\\Roaming\\jupyter\\runtime\\kernel-1234.json' This will work either for Linux or Windows.

  3. Start the remote client:

    3.1. Once you locate the file, copy it over to your server machine using scp in linux or pscp or winscp in windows SCP w/ ssh: copying a local file from windows to a remote server using scp

    3.2. Make sure that you are the same directory as the kernel-1234.json file.

    3.3. Open up the kernel-1234.json file using vim or your favorite text editor. You will notice a line saying "ip": "127.0.0.1". Change 127.0.0.1 the ip address from your server that is visible from the client, that you found in step 1.3 and save the json file.

    3.4. Start up the remote kernel using jupyter console –existing=kernel-1234.json, while located in the same drive where kernel-1234.json is located.

If you have followed the steps above, you should now be able to connect to the remote ipython kernel regardless of whether the ipython kernel is running on Windows/Linux/Mac.

alpha_989
  • 4,882
  • 2
  • 37
  • 48
1

I tried the manual way on https://github.com/ipython/ipython/wiki/Cookbook%3a-Connecting-to-a-remote-kernel-via-ssh once again and it worked. In detail:

windows-machine$ ipython kernel -f kernel-1234.json

linux-machine$ scp windows-machine:path/to/kernel-1234.json .
linux-machine$ cat kernel-1234.json
{
  "stdin_port": 55534, 
  "ip": "127.0.0.1", 
  "control_port": 58391, 
  "hb_port": 60540, 
  "signature_scheme": "hmac-sha256", 
  "key": "fa461cf7-f078-4c22-909f-cfa7d2a30596", 
  "shell_port": 60159, 
  "transport": "tcp", 
  "iopub_port": 59207
}
linux-machine$ ssh -f -N -L 55534:127.0.0.1:55534
linux-machine$ ssh -f -N -L 58391:127.0.0.1:58391
linux-machine$ ssh -f -N -L 60540:127.0.0.1:60540
linux-machine$ ssh -f -N -L 60159:127.0.0.1:60159
linux-machine$ ssh -f -N -L 59207:127.0.0.1:59207
linux-machine$ ipython console --existing kernel-1234.json
Anders Lundstedt
  • 131
  • 9