Re: CouchDB High Availablility deployment and ChangeFeed

[Jan Lehnardt <ja...@apache.org>]

Hi A,

for your setup, we recommend using CouchDB Clustering to achieve high availability.

But note that neither a single node nor a cluster _changes feed guarantees exactly-once-delivery. You have to solve that in your Spring-boot API.

Best
Jan
—
Professional Support for Apache CouchDB:
https://neighbourhood.ie/couchdb-support/

*24/7 Observation for your CouchDB Instances:
https://opservatory.app

> On 27. Mar 2022, at 13:45, Al Z. <az...@gmail.com> wrote:
> 
> Hello.
> 
> I have a single node running couchDB 3 deployed live.
> I have a DB-per-user setup with about 2k databases on a single node.
> A single Spring-boot API talks to couchDB vis HAProxy ...
> 
> 1) I would like to have multiple 2 to 3 nodes in the same DC so that in
> case one node goes down, the system is still operational.
> Each node will have an instance of the Spring-boot API, ...
> I am not sure whether I should be looking into clustering or replication.
> I do understand that with replication, the nodes can be in different DB.
> At this point, the plan is a single DC only. In the future, I may look into
> a cross DC setup.
> 
> Any hint will me most appreciated.
> 
> 
> 2) I use the couchDB change feed to update some internal systems and send
> notification emails.
> Each change feed needs to be processed exactly once.
> The change feed seems to be a major point of failure for my setup ...( I
> also noticed that if couchDB is restarted, I need to restart the
> spring-boot app, otherwise, it does not process the feed).
> Given that I am planning to move to a multiple nodes setup, I am not very
> sure how I would be able to process each feed only once without having to
> manually handle duplicate.
> Any experience or hint or alternative would be most appreciated.
> 
> Thank you very much.
> 
> Best regards.
> 
> A.