[
https://issues.apache.org/jira/browse/CASSANDRA-2306?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13005041#comment-13005041
]
Sylvain Lebresne commented on CASSANDRA-2306:
---------------------------------------------
bq. I don't think so. I think it's better to fail a read command with QUORUM or
ALL CL if there is no enough live node in a local DC. If a read command need to
get data from a node from another DC performance degradation will be
unacceptable in many situation because of network latency. (This is true to
write performance also. LOCAL_QUORUM performs much better than EACH_QUORUM.)
There are a lot of situation where people are very happy with getting that data
even if it has to cross the atlantic than getting no data at all. And that's
not a supposition. Hence forbidding this is a serious drawback for Cassandra.
And anyway, LOCAL_QUORUM does exactly this in a by-request configurable way,
which is way better.
> generate endpointList depending on command type (read or write)
> ---------------------------------------------------------------
>
> Key: CASSANDRA-2306
> URL: https://issues.apache.org/jira/browse/CASSANDRA-2306
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Reporter: ivan
> Priority: Minor
> Attachments: rw_sep.patch
>
>
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
