Is it the right way to load balance couchdb with nginx?

[estellederrien]

So i have this 3 machine couchdb cluster working very well . Each are ubuntu 18.0 versions.

So now, my goal is to add a 4th machine, with NGINX as a load balancer .

I 've edited my following file nginx-config-directory/sites-available/defaut

And have added this :

upstream couchDbCluster {
   server 10.1.0.101:5984; 
   server 10.1.0.102:5984;
   server 10.1.0.103:5984;
}


location /couchdb {
    rewrite /couchdb/(.*) /$1 break;
    proxy_pass http://couchDbCluser;
    proxy_redirect off;
    proxy_buffering off;
    proxy_set_header Host $host;
    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
}

Do you think it is the right syntax ?

i wonder if one gentle person could tell me if that is the right way to load balance inside my couchdb ?

What is strange is that i would like to separate when it is a GET or a PUT for example, and this code doesn't seem to care about this . How could i do faster GET queries ?

Do I have to change another file ?

Thanks you a lot for your help, have a nice day !

[wohali]

Through experience, CouchDB developers recommend using HAProxy as a load balancer. We provide a default configuration that just works™ here: https://github.com/apache/couchdb/blob/master/rel/haproxy.cfg

We do document our recommended Nginx configuration as well. Getting it right is difficult, and there are specific issues with configuring Nginx properly as a reverse proxy for CouchDB when you are using replication, so be sure to review our documentation.

At present, CouchDB does not natively have the concept of "read-only replicas," so there is no need to separate GET and PUT requests.

Finally, remember that in a 2.x CouchDB cluster, additional nodes don't participate in requests for databases you've already created unless you move shards to those new nodes. See the new 2.2.0 documentation on shard management for more details.