Re: 1.1.0 replication specifics

[Randall Leeds <ra...@gmail.com>]

Search on this mailing list. I've outlined some of the replication specifics
before here in response to similar questions.
I'll make a note on my todo to write some of it up more formally somewhere.

On Tue, Aug 30, 2011 at 06:39, Eric Abbott <er...@viximo.com> wrote:

> I've pulled down and am playing with 1.1.0 and have a few questions about
> replication.
> First off, are there any in-depth documents out there describing
> replication
> and/or use case scenarios.
>
> I'm looking for detail to explain the following:
> [One quick note on all this, the 25Gb data file is couch v1.0.2. v1.1.0 was
> installed and the data file kept]
> *) Bulk loading: is /db/_bulk_docs really the only option?
>  --> Using _bulk_docs in 10k chunks, I end up with a 25Gb file for the
> original 3.3Gb json data of 17mil records.
> *) Adding a new server: replication
>  --> Copy db file over, start new server. add pull replication back to
> original server.
>       Replication takes 5-8hrs to ascertain there are no differences
>       * Not practical with 5 multi-masters and each pass takes 5+ hours
>       * Poking around the source code, I see "find_replication_log()".
> Whats this?
> *) Error state: Replication got into an error state at or near the end of
> replicating 17mil rows
> curl http://localhost:5983/_replicator/pull3
> {"_id":"pull3","_rev":"3-7e30479f975007bd9a08c321a124ac9e","source":"
> http://ec2-50-16-182-193.compute-1.amazonaws.com:5983/demographics
>
> ","target":"demographics","continuous":true,"_replication_state":"error","_replication_state_time":"2011-08-30T12:51:01+00:00","_replication_id":"c9dffd968a3e614ba344128e7534b67a"}
>  --> How do we kick start out of this state? Restarting the couch servers
> on both ends had no effect.
>