Connecting to production node when short name is set

Question

I am running an Elixir app on my 3 servers. I recently needed to connect nodes so that I can sync something between them, and for that I set short names for each node.

I am using Edeliver/Distillery to generate release and deploy. I copied var/vm.args to release directory on each node and changed -name blah@ip to -sname name@nodename. Nodes can see and connect to each other without problem.

Before I given each node a name, I used ssh to remotely connect to my nodes when I needed to debug my app. I used this approach:

• ssh node -L 4369:127.0.0.1:4369
• run epmd -names on my local machine to find out port of my app
• kill last ssh and run ssh node -L 4369:127.0.0.1:4369 -L port:127.0.0.1:port to have access to running node on my machine
• run iex --hidden --erl '-name debug -setcookie cookie_like_server' on my machine
• now I had access to remote node, e.g. I could use observer to see my remote node

The same procedure won't work now.

First I see an error saying nodes using short and long names cannot connect to each other, obviously.

I changed my iex command to iex --hidden --erl '-sname debug -setcookie cookie_like_server' to use short name. This way connection times out when I try to connect to remote node on observer.

I don't really get the problem. Am I doing something wrong? Is here another way to get access to a running node that uses short name?

BTW, I used ssh so that my traffic will be encrypted. Please do tell me if there is a solution that makes my connection non-encrypted, though I will probably not use it, or somehow tunnel it through ssh.

UPDATE

I tried method suggested by @christophe-de-troyer. TLDR, didn't work.

vahid@arch-adtube ~/ % iex --sname debug --cookie 'cookie' --remsh backend-platform@prod-1
Erlang/OTP 20 [erts-9.0.1] [source] [64-bit] [smp:4:4] [ds:4:4:10] [async-threads:10] [hipe] [kernel-poll:false]

Could not contact remote node backend-platform@prod-1, reason: :nodedown. Aborting...

ssh prod-1 -L 4369:127.0.0.1:4369 was running, and epmd -names reports backend-platform as a running node correctly.

Answer 1

If you want to connect to a remote node you will have to pass a full name (--name) instead of a short name (--sname), from my experience.

So on Node A, with ip 192.168.1.10, you run:

iex --name [email protected] --cookie secret

On node B, with ip 192.168.1.11, you run:

iex --name [email protected] --cookie secret

And then on your computer, with ip 192.168.1.12, you can attach to these remote shells using. First of all you have to make sure that your local EPMD port (4369) is bound to the remote server's EPMD port, by tunneling it over SSH.

ssh a -L 4369:127.0.0.1:4369

Then you can connect with a remote shell to these nodes using:

iex --name [email protected] --cookie secret --remsh [email protected]

I have experimented a bit on my machines, and I found out that it only works with IP addresses. My machine running at machine.example.com, with hostname machine, was not connectable if the name on the remote machine was not qualified with its ip address, or domain name. And I had to match that with the DNS name, or ip address on my connecting client.

There is some interesting traffic on the mailing list of Erlang about this[1].

[1] http://erlang.org/pipermail/erlang-questions/2006-December/024270.html

Edit

I did some further experimentation and succeeded in connecting to remote nodes with shortnames.

As you already mentioned, you indeed need the two nodes to be running in the same naming mode (i.e., both short, or both fully qualified).

On node A, with ip 12.12.12.12, and FQDN a.example.com, I spawned a shell as such:

iex --sname bob --cookie secret

On my local machine L, with ip 5.5.5.5, and FQDN l.notanexample.com, I spawned a shell as such:

ssh a -L 4369:127.0.0.1:4369

iex --sname alice --cookie secret --remsh bob@a

The reason this works, and did not before, was because on my local machine, in the /etc/hosts file, I did not put a line that bound the hostname (short hostname) to the remote IP.

So in /etc/hosts on my local machine I have the following line now:

12.12.12.12 a.