Documentation Lacking around Routing Key Implications about rabbitmq-consistent-hash-exchange HOT 4 CLOSED

Any information on this would be helpful.

from rabbitmq-consistent-hash-exchange.

My 2 cents of feedback, related to the (multiple) requests of @richbria.

In regards to improving the documentation of the routing keys usage, I would suggest adding some link to a consistent-hashing-for-dummies blog post, there are some of them quite comprehensive (example or example). Also, referring to the concept of points as replicas instead/as well, might help understand the usage of the routing keys quicker.

+1 that it would be nice to have a recommendation on the magnitude of points/replicas to use based on the performance implications of the current implementation.

+1 that if would be nice being able to shard based on a segment of the routing key.

from rabbitmq-consistent-hash-exchange.

@arnauorriols I will add your blog post links but consider starting a rabbitmq-users discussion instead of posting +1s here. The problem with +1s is that they rarely lack the important technical details that engineers need, e.g. how do you define "shard" and "segment of the routing key". Thank you.

from rabbitmq-consistent-hash-exchange.

With a single consumer it doesn't matter because the entire hashing ring is allocated to a single "node" (queue in this case). Consistent hashing is based on a range of integer values, not percentage of the ring, historically, and changing that in this plugin would be a breaking change that affects every existing user.

Some understanding of what consistent hashing is is necessary when using this plugin. For those looking to partition a stream of messages across N consumers in a roughly even manner while also trading off total ordering, rabbitmq-sharding would probably be an easier to understand option.

from rabbitmq-consistent-hash-exchange.