1-to-1 connections & possibly hidden nodes

[Michael Truog]

On 7/12/21 3:02 PM, Roberto Ostinelli wrote:
>
> What is the advantage over using a hidden node? Here below is the 
> output with one@ and two@ in a full cluster, while three@ has been 
> started with the -hidden option and has been connected only to one@:
>
> (*one*@dev.local)1> nodes().
> ['two@REDACTED']
> (*one*@dev.local)2> nodes(hidden).
> ['three@REDACTED']
> (*one*@dev.local)3> register(shell, self()).
> true
>
> (*two*@dev.local)1> nodes().
> ['one@REDACTED']
> (*two*@dev.local)2> nodes(hidden).
> []
>
> (*three*@dev.local)1> {shell, 'one@REDACTED'} ! hello.
> hello
>
> (*one*@dev.local)4> flush().
> Shell got hello
> ok
>
>
Hi Roberto,

It is best if you have the kernel app 'dist_auto_connect' setting set to 
'never' if you want to ensure each node connection is created 
programmatically (to avoid unintended node connections due to 
messages).  The function net_kernel:connect_node/1 is creating a visible 
node connection, ensuring that your connection makes your node part of 
the fully connected set of visible nodes.  If you used 
net_kernel:hidden_connect_node/1 a hidden node connection would be 
created and you will not get any additional node connections from that 
function call.  However, net_kernel:hidden_connect_node/1 remains 
undocumented so it isn't clearly part of the public interface.  CloudI 
(https://cloudi.org) uses it and it is the preferred function for 
connecting to an Erlang node from a script, so I would like to think the 
function isn't disappearing in the future.

Best Regards,
Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://erlang.org/pipermail/erlang-questions/attachments/20210712/d941a3ee/attachment.htm>