I had noticed when stepping through the manual/API setup process that
these databases were being created implicitly by one of the steps, and
it seems that it's the `finish_cluster` action that does this.

This answers my second question regarding the 'config-only' approach -
it seems there does need to be an additional step to call
`finish_cluster`, aka create the 'cluster system DBs'.

Thanks for clarifying Joan.

--
John

John Fitzpatrick
[email protected]

Datalink Internet Systems Pty Ltd
Main: www.datalink.com.au • 1300 DATALINK
Direct: +61 475 421 064


On Tue, 22 Jan 2019 at 11:45, Joan Touzet <[email protected]> wrote:
>
> Hi John,
>
> The _cluster_setup endpoint checks for the presence of _users,
> _replicator and _global_changes as clustered databases.
>
> The seedlist setup only makes copies of the local DBs, which are
> _dbs, _nodes and _users, across the entire cluster.
>
> You still need to create the _replicator DB, and _global_changes if
> desired. (It is, of course, possible to run a cluster without either
> of these if desired, or to use a custom replicator DB, which is any
> DB that ends in /_replicator.)
>
> -Joan
>
> ----- Original Message -----
> > From: "John Fitzpatrick" <[email protected]>
> > To: [email protected]
> > Sent: Monday, January 21, 2019 7:15:57 PM
> > Subject: Is 'finish_cluster' needed when using 'seedlist' config?
> >
> > Hey CouchDBers,
> >
> > I have just worked through the process of creating a cluster using
> > the
> > setup API, and have since discovered the cluster `seedlist` config
> > feature introduced in 2.3.0.
> >
> > This option makes the setup process much more straightforward,
> > however
> > I've noticed that, on a seeded cluster, the `_cluster_setup` endpoint
> > still returns `cluster_enabled` rather than `cluster_finished`.
> >
> > Is it the intention that the `finish_cluster` action still needs to
> > be
> > invoked on the cluster setup API? Or is there a way to bootstrap a
> > cluster using a 'config-only' approach?
> >
> > Regards,
> > --
> > John Fitzpatrick
> > [email protected]
> >

Reply via email to