Mailing-List: contact dev-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@couchdb.apache.org
Message-ID: <1976870648.1254414923850.JavaMail.jira@brutus>
Date: Thu, 1 Oct 2009 09:35:23 -0700 (PDT)
From: "Ning Tan (JIRA)" <jira@apache.org>
To: dev@couchdb.apache.org
Subject: [jira] Created: (COUCHDB-516) Replication: _replicate does not
 finish replication in one pass and has to be invoked repeatedly
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

Replication: _replicate does not finish replication in one pass and has to be invoked repeatedly
------------------------------------------------------------------------------------------------

                 Key: COUCHDB-516
                 URL: https://issues.apache.org/jira/browse/COUCHDB-516
             Project: CouchDB
          Issue Type: Bug
          Components: Database Core
    Affects Versions: 0.10
         Environment: Mac OS X 10.5 w/ CouchDB 0.10.x multiple revisions (latest is 820436)
Ubuntu Desktop 9.04 w/ CouchDB 0.10.x multiple revisions (latest is 818247)
            Reporter: Ning Tan


When we replicate between a remote database and a local one (pulling from remote into local), we are observing partial replications, meaning that we have to issue repeated _replicate calls for the replication to complete. For a database with 17,000 documents, for example, it could take up to 7 calls for the entire database to replicate into an empty one. Each time, the number of documents replicated over seemed random.

The use case is very simple--replicating a database into an empty one with no concurrent writes, no additional load or i/o, etc. The databases involved are a mixture of the 10.0.x code base natively built on Ubuntu and Mac.

It seems to me that every (not all) partial replication process is associated with a corresponding entry in the log that says "recording a checkpoint at source update_seq .....". (i.e. you can match the recorded_seq number in the replication response with the checkpoint update_seq numbers in the log).

Futon vs. curl doesn't make a difference.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.