Question about lifetime

[bjtaylor1]

Hello, this is more of a question than an issue as such, but I would like to know please about how to use the amqp client in a long running service.
This has a function that has called regularly on a schedule within the scope of a node.js program, which is run as a service (on ubuntu).
Is it advisable/more efficient to declare the AMQPClient and AMQPChannel in global scope, meaning I only have to connect once at the start of the program, and use them in the function that is called regularly, OR, is it better to declare a new client and channel inside the function?

If the former, is there a way of determining whether the channel is still connected, and reconnecting if not?
Or if the connection drops, does it have the ability to reconnect?

If the latter, is there any inefficiency associated with keep reconnecting again repeatedly?

[dentarg]

It is good practice to re-use the client connection and channel to avoid connection churn (see more at https://www.rabbitmq.com/docs/connections).

However, it is also good practice (important) to separate publishing and consuming, see https://www.rabbitmq.com/docs/channels#flow-control

Lastly, if you do need to re-connect repeatedly (like on every request to your web application), you could consider using https://github.com/cloudamqp/amqproxy