[jira] [Commented] (COUCHDB-2822) Replication not replicating every document

["Panu Horsmalahti (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/COUCHDB-2822?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14910120#comment-14910120 ] 

Panu Horsmalahti commented on COUCHDB-2822:
-------------------------------------------

The issue might be caused by the target document having "_rev": "102-e83ff1aefa1bcd7165cc9e8b7e14532d" while the source document has "_rev": "18-a3159727cf84daf13992d181d25e83de" even though the source document should containt more updates than the target document. I don't understand why the target document has such a high revision number. I wonder if npm changed the revision numbers at some point.

> Replication not replicating every document
> ------------------------------------------
>
>                 Key: COUCHDB-2822
>                 URL: https://issues.apache.org/jira/browse/COUCHDB-2822
>             Project: CouchDB
>          Issue Type: Bug
>      Security Level: public(Regular issues) 
>          Components: Replication
>            Reporter: Panu Horsmalahti
>
> I'm replicating the NPM's skimdb from https://skimdb.npmjs.com/registry/ to my own, private database. The replication has been working for about a year, but it was timed out for a month. Now if I start the replication again, it's successfully picking up all the new changes to the database, but not replicating any (or almost none?) of the changes for the duration of the downtime of the replication.
> Here's the replicator document:
> {code}
> {
>    "_id": "registry",
>    "_rev": "193-8a190eb314c9ee901eaa4127db02a05b",
>    "source": "https://skimdb.npmjs.com/registry",
>    "target": "registry",
>    "user_ctx": {
>        "name": "root",
>        "roles": [
>            "_admin"
>        ]
>    },
>    "continuous": true,
>    "owner": "root",
>    "proxy": "<censored>",
>    "create_target": true,
>    "_replication_state": "triggered",
>    "_replication_state_time": "2015-09-22T09:46:32+03:00",
>    "_replication_id": "f357bfe5f63bb9daed6e55e6223ee1b5"
> }
> {code}
> The replication status is:
> Replication	https://skimdb.npmjs.com/registry/ to registry	2015-09-22 10:27:26	2015-09-22 10:43:14	<0.148.0>	Checkpointed source sequence 307921, current source sequence 307921, progress 100%
> I can manually check a document which has not been updated to my own database, for example the "nan" document. I can see it being changed here: https://skimdb.npmjs.com/registry/_changes?descending=true&limit=50000
> {code}
> {"seq":277249,"id":"nan","changes":[{"rev":"18-a3159727cf84daf13992d181d25e83de"}]},
> {code}
> That revision is indeed the latest revision in the source database. The document in my database has a different revision id.
> I've tried restarting CouchDB multiple times and recreating the replication document. I also tried to create a one-shot (instead of continuous) replication while specifying only this particular document, but it was reported that there was nothing to be replicated (no missing revs).
> While the database in question is NPM specific, I don't see how this could be a NPM issue.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)