NATS: First Impressions | Hacker News Reader

NATS: First Impressions

blog.sequin.io

24 points_acco7 comments

Open articleView on HN

https://nats.io/ (Tracker removed)

> Connective Technology for Adaptive Edge & Distributed Systems

> An Introduction to NATS - The first screencast

I guess I don't need to know what it is

The first paragraph in the article explains it. It is similar to RabbitMQ.

> NATS is an open-source messaging system. At its core, NATS provides publish-subscribe messaging, but has expanded to enable other features. It has a request/reply mechanism which you can use for service-to-service communication (in lieu of HTTP APIs). And it now includes a persistence layer which means it can on workloads you'd normally turn to Kafka for. The project is supported by the CNCF.

Does according to author, NATS is less reliable, if you want guaranteed ordering and to have multiple instances of worker, because NATS partitions lack "auto-balance" like in Kafka?

I would assume that NATS can be deployed in some Kubernetes cluster, so it would not be uncommon to have 3 or 5 workers. In this case, what would happen, if someone wanted guaranteed ordering and used that deterministic subject token partitioning?

Let's assume that some worker would crash due to lack of storage space and it would keep restarting, so what happens to messages on this partition? Can they be processed, if specific worker is not available? Is it possible to react to this event and manually reassign these partitions to other workers? In case that there's no event, then maybe it's possible to write some script and run CronJob to manually check, if rebalance is needed?

A feature not covered in this article that I recently learned about is republishing[0]. This is super useful, because it lets you essentially "namespace" parts of the network. You can use very short subject names for an isolated part of the network, and then republish to a more specific name as it gets pushed out to a larger network.

[0] https://docs.nats.io/nats-concepts/jetstream/streams#republi...

Here in my company we're running load tests against a POC application using NATS, and we're afraid that it can't handle some spikes in load that well. In some tests rounds the loss rate was a bit high. But I can't say if it's a NATS issue or our tests aren't well dimensioned.

If you were writing your own NATS client from scratch, this is how you'd publish a message: ...

Not how I would do it.

I don't know Nats, could you explain briefly how then?

Reply on news.ycombinator.com