couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Anderson (JIRA)" <j...@apache.org>
Subject [jira] Commented: (COUCHDB-686) Improve ordering of replication
Date Wed, 10 Mar 2010 17:57:27 GMT

    [ https://issues.apache.org/jira/browse/COUCHDB-686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12843665#action_12843665
] 

Chris Anderson commented on COUCHDB-686:
----------------------------------------

Randall's suggestion seems the most appropriate here. However, I'd caution against this using
such a feature. For the same reasons we don't support multi-document transactions on a single
node (b/c they can't be sanely supported on a cluster), supporting ordering guarantees is
dangerous. They are essentially a form of multi-document state.

Since is is very clear that they can't be supported in a replication mesh with more than 2
peers, I don't think they are so dangerous as that we shouldn't provide Randall's knob. But
we should be clear to the people who choose to architect around it, that what they are doing
will be limited to the maximum throughput of a single disk.

> Improve ordering of replication
> -------------------------------
>
>                 Key: COUCHDB-686
>                 URL: https://issues.apache.org/jira/browse/COUCHDB-686
>             Project: CouchDB
>          Issue Type: Improvement
>          Components: Replication
>    Affects Versions: 0.10.1
>            Reporter: Dirkjan Ochtman
>            Priority: Minor
>             Fix For: 0.12
>
>
> Currently, (continuous) replication almost preserves the ordering of the updates. It
would be nice if this could be fixed so that the order of outgoing updates fully corresponds
to the order inside the source database. While this may fall down in some clustering situations,
it's a great guarantee to provide in many smaller situations involving an ordered stream of
documents/updates.

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


Mime
View raw message