incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Robert Newson <rnew...@apache.org>
Subject Re: when replication hits wrong update_validation
Date Tue, 02 Jul 2013 19:40:26 GMT
I believe it's "since_seq". Try adding "since_seq":0 to your POST body
to _replicate.

B.


On 2 July 2013 20:36, Robert Newson <rnewson@apache.org> wrote:
> It suffices to delete the checkpoint document, though I'm sure there's
> a not-well known option for force a sequence. I'll hunt for it.
>
> B.
>
>
> On 2 July 2013 20:16, Jens Alfke <jens@couchbase.com> wrote:
>>
>> On Tue, Jul 2, 2013 at 5:44 AM, svilen <az@svilendobrev.com> wrote:
>>
>>> a) can i force it to copy all again? i'm not sure if that "since" number
>>> is kept in touchdb-localy or on server too.
>>
>> The checkpoint is kept on both sides to ensure consistency. I think you’ll need
to delete and recreate one of the databases (either on device or CouchDB) to reset the replication.
On iOS there is a “reset” property you can set in the replication spec to force starting
over, but I don’t think that’s implemented on Android yet.
>>
>>> b) how is this case handled in plain couchdb doing replication?
>>
>> The replicator isn’t supposed to treat a document as having been replicated unless
the PUT either succeeded or returned a 403 Forbidden status. The 500 error you got in this
case should have left the local doc’s status as unsynchronized, so the replicator would
try it again next time.
>>
>> Tracking the state of which docs have been synced and which haven’t is pretty tricky.
In the case of TouchDB / Couchbase Lite, here were a bunch of improvements made on iOS that
haven’t been ported to Android yet.
>>
>> —Jens

Mime
View raw message