The setup endpoint was easy to use and I have a lot more confidence that
I've set it up correctly, thanks.

The only thing that threw me was the endpoint name and json format are
incorrect in the README. I was going to fork and fix it but it looks like already did pretty much
every change I was going to make. His stuff just hasn't made it into the

Thanks for your help,

On Thu, Oct 13, 2016 at 4:11 AM Jan Lehnardt <> wrote:

> On 12 Oct 2016, at 22:40, Timothy McKernan <>
> You are right, I did skip over it. I apologize for the easy miss.
> If you'll bear with me, here is what I was seeing, and why I skipped over
> it: I only use ssl and was trying to automate that step (for an rpm
> installation). The behavior seems different under ssl than port 5984. My
> bit adds the admin user and binds to easily enough. But when I run
> setup and am viewing fauxton under https the list of db's shows missing
> shards, such as this:
> shards/00000000-1fffffff/_global_changes.1476302646

shards don’t show up on 5984, they are an implementation detail that is
(for the moment) only relevant on a per-node basis (and thus 5986).

> And likewise there are lots of errors in the logs:
> [error] 2016-10-12T20:22:01.293892Z couchdb@localhost <0.9859.0> --------
> httpd 400 error response: {"error":"illegal_database_name","reason":"Name:
> 'shards%2F60000000-7fffffff%2F_users.1476302646'. Only lowercase
> (a-z), digits (0-9), and any of the characters _, $, (, ), +, -, and / are
> allowed. Must begin with a letter."}
> I'm really only explaining all of this to get around to my next question:
> Those shards/errors that are created during setup don't appear under
> fauxton:5984, but they do under fauxton:5986 and fauxton:6984. Am I
> something in my ssl setup?

Yes, you need to run the equivalent of creating the three databases:
_users and _replicator.

Or use the /_setup_cluster endpoint as documented (roughly) here:

It is what Fauxton uses under the hood, so you can script it.

* * *

That said, we totally acknowledge that this is a bit confusing and
under-documented, so we hope you bear with us :)


> On Wed, Oct 12, 2016 at 3:35 PM Jan Lehnardt <> wrote:
>> Looks like you skipped over the "setup cluster" step from the
>> notes on Even a single/node cluster-of-one needs this,
>> as it creates the necessary system databases.
>> At this point, best to ignore the 5986 port (unless for per-node
>> maintenance).
>> Cheers
>> Jan
>> --
>>> On 12 Oct 2016, at 20:44, Timothy McKernan <>
>> wrote:
>>> I've installed CouchDB 2.0.0 from source using:
>>> ./configure
>>> make release
>>> sudo mv rel/couchdb /var/lib/couchdb
>>> sudo chown -R couchdb:couchdb /var/lib/couchdb
>>> sudo -u couchdb /var/lib/couchdb/bin/couchdb
>>> The problem I'm seeing is that, by default, [chttpd] is being assigned
>> port
>>> 5984 (in the default.ini) and [httpd] is assigned 5986.
>>> Throughout the online docs port 5984 is referenced for the majority of
>>> endpoints, such as fauxton.
>>>   Used for pretty much everything, such as couchapps (fauxton is the new
>>> futon) and db access.
>>> And port 5986 is the "back-door" port.
>>> So we end up with the following behavior:
>>> md-007:~ $ curl -k -X GET
>>> []
>>> md-007:~ $ curl -k -X GET
>>> ["_dbs","_nodes","_replicator","_users","md_rules","monitor"]
>>> md-007:~ $ curl -k -X GET
>>> ["_dbs","_nodes","_replicator","_users","md_rules","monitor"]
>>> It seems that chttpd doesn't have the usual default access to everything
>>> that it used to in couchdb 1.6. Indeed, if you use fauxton over port
>>> as the docs say to, it _doesn't_ show you any db's, because it clearly
>>> (from the above curl examples) can't see any. And the emulator complains
>>> because it's using 5984 to try and find _users.
>>> [error] 2016-10-12T18:42:53.106630Z couchdb@localhost emulator --------
>>> Error in process <0.4185.0> on node couchdb@localhost with exit value:
>>> What's the correct configuration here? What ports should chttpd and
>>> be set to so that the emulator process isn't erroring?

Professional Support for Apache CouchDB:

Reply via email to