cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benedict (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-8990) Allow clients to override the DCs the data gets sent to, per write request, overriding keyspace settings
Date Wed, 18 Mar 2015 19:57:39 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-8990?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14367775#comment-14367775
] 

Benedict commented on CASSANDRA-8990:
-------------------------------------

This doesn't seem likely to ever be introduced, as it would never be clear to a query where
the data to answer the query should be requested from. Unless you provided this information
on query as well, but this is very burdensome, likely error prone, and unlikely to be used
widely, so not really worth the effort to support.

Why is it a problem to have multiple keyspaces for different replication criteria?

You can, as [~philipthompson] says, accept fewer data centres / nodes receiving the data at
write time than you _require_ - but the data will make it to the left out nodes eventually
(and typically very quickly)

> Allow clients to override the DCs the data gets sent to, per write request, overriding
keyspace settings
> --------------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-8990
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-8990
>             Project: Cassandra
>          Issue Type: New Feature
>            Reporter: Donald Smith
>
> Currently each keyspace specifies how many replicas to write to each data center. In
CQL one specifies:
> {noformat}
>  WITH replication = {
>   'class': 'NetworkTopologyStrategy',
>   'DC1: '3',
>   'DC2': '3'
> }
> {noformat}
> But in some use cases there's no need to write certain rows to a certain datacenter.
 Requiring the user to create two keyspaces is burdensome and complicates code and queries.
 
> For example, we have global replication of our data to multiple continents. But we want
the option to send only certain rows globally with certain values for certain columns -- e.g.,
only for users that visited that country).
> Cassandra and CQL should support the ability of client code to specify, on a per request
basis, that a write should go only to specified data centers. 



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

Mime
View raw message