opening Publishers via lookupd?

[silmeth]

Is there any plan to implement features similar to those of Subscriber for a publishing client, ie. make something Publisher-like that would be able to connect via lookup, allow automatic reconnects, etc.?

As I understand the documentation, at the moment if I want to have a program both publishing on some topics and listening on others, then on one hand I need to create a Subscriber which will manage connections to nsqd instances automatically via provided lookup addresses (via a list of NsqLookupDHTTPClient instances), but on the other hand I need to open my own separate list of lookup clients to discover instances for publishing and manage those connections myself?

That sounds like a lot of unnecessary code duplication to manage the lookup logic, is there a deeper reason for not factoring the lookup logic out of Subscriber and sharing it between subscriptions and publishers? Or perhaps just making Subscriber able to spawn both Subscriptions and Publishers?

I’m not that well familiar with NSQ so I hope what I wrote above makes sense.

[Chrzi]

The underlying design idea for NSQ is to have the producers and the nsqd instances run together on one machine. A lookup mechanism for the producers is therefor not necessary, as each producer will publish to its local instance. This setup is described here in the docs: https://nsq.io/deployment/topology_patterns.html

That is also reflected in the official libraries and we try to have the same kind of API interface as them.

There are reasons to a different setup, so I am not opposed to adding something like that.

[silmeth]

Thanks for the response! Yeah, after writing this question, I looked into some other clients and found similar discussions there and realized that this is by design and an something many users comment on (I guess the thing feels unintuitive to many folks, me included, while not being communicated enough in the documentation).

I previously had contact with one library which did allow creating producers via lookup instance (I guess it just listed all nodes and chose randomly from those with created topic), so didn’t even occur to me this is not the intended usage pattern.

Since this is by design, I’m fine with it staying as it is – but I (and I guess other NSQ newbies) would appreciate a few additional sentences in the docs stating rationale for the limitation.