Creating a new NATS connection per each subscriber

[vedadiyan]

I am working on creating a fully automated microservice framework and I found that if I create a new connection for each subscriber I may get the following benefits:

• I can name the connection and I can get out-of-the-box health monitoring through the built-in monitoring endpoint
• I can simplify service binding by eliminating global variables. This is useful if two microservices would need to connect to two different servers, however this is a very rare case.

However, my concern is:

• In case there would be too many subscribers and in case each subscriber runs in cluster mode by being attached to a queue, would it cause port exhaustion in the server?
• Will the server get slow due to too many connection?

In my opinion this approach has it's pros and cons, but I need to clarify this so I really appreciate if someone helps me better analyze this case to make sure it is production wise.

[derekcollison]

We would generally not suggest a connect per subscriber. However sometimes this becomes a natural pattern if an app has one main subscription.

When you ask the system about connections via the nats cli, you can filter those connections based on a subject. So you could monitor all connections that have a subscription to say foo.bar.

[vedadiyan]

Thanks