Creating a 2-node cluster with couchdb 2.1 never works

[harald-d]

I tried to create a 2-node cluster multiple times (both on windows and ubuntu) and it never worked. I did exactly as described in the official documentation here:
http://docs.couchdb.org/en/latest/cluster/setup.html#the-cluster-setup-api

Expected Behavior

When i finish the cluster-configuration of the two nodes, i expect to create a database on node1, which should show up on node2. Verification on both nodes via fauxton fails also "internal server error" - that happens both under linux and windows with version 2.1

Current Behavior

I configure both CouchDB via API

node1: (10.0.0.1)
1. POST {"action": "enable_cluster", "bind_address":"0.0.0.0", "username": "admin", "password":"mypassword", "node_count":"2"}

2. POST {"action": "enable_cluster", "bind_address":"0.0.0.0", "username": "admin", "password":"mypassword", "port": 5984, "node_count": "2", "remote_node": "10.0.0.2", "remote_current_user": "admin", "remote_current_password": "mypassword" }

3. POST {"action": "add_node", "host":"10.0.0.2", "port": "5984", "username": "admin", "password":"mypassword"}

4. POST {"action": "finish_cluster"}

Membership-Status:
{"all_nodes":["couchdb@localhost"],"cluster_nodes":["couchdb@10.0.0.2","couchdb@localhost"]}

node2(10.0.0.2)
same configuration as node1, but IP address for other node changes to 10.0.0.1

Membership-Status:
{"all_nodes":["couchdb@localhost"],"cluster_nodes":["couchdb@10.0.0.1","couchdb@localhost"]}

I never defined a zone - is this mandatory for the cluster to work ?
Is there another recommended way to test if the cluster is working ? The goal is to create a 2-node cluster where one node can fail and the dbs continue to operate via node2.

Your Environment

Tried on Windows 10, Windows Server 2016, Ubuntu 14.04
CouchDB 2.1

[wohali]

First, update to 2.1.1, this is a critical security patch.

Then, you have to change your node names in vm.args. You seem to have left the 2.1.0 default of -name couchdb@localhost. Read the comment carefully. In your situation, these will need to read something like -name couchdb@10.0.0.1 and -name couchdb@10.0.0.2.

Once you've made these changes, restart the nodes and try your setup again. You only have to perform the POSTs on a single node, no need to repeat those steps on the other node.

Finally, as a reminder, CouchDB clustering works best in a 3-node cluster.

[harald-d]

Hello Joan, I can't thank you enough - changing the name in vm.args did the trick, now the cluster seems to work. (Maybe someone adds this to the documentation because either i overlooked it or it isn't documented) Thank you very much for your help and have a nice day. best regards, Harald Joan Touzet <notifications@github.com> schrieb am Fr., 10. Nov. 2017 um 18:43 Uhr:

…

Closed #978 <#978>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#978 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/Af_QP7qOhcCz6z9jJ7vN2uHE17PeyL45ks5s1Is_gaJpZM4QZSCn> .

[wohali]

You're welcome!

It's in the documentation a bit farther up the page:

http://docs.couchdb.org/en/latest/cluster/setup.html#firewall

but I agree the docs could be improved.