[ https://issues.apache.org/jira/browse/CASSANDRA-6124?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13786586#comment-13786586 ]
Patricio Echague commented on CASSANDRA-6124: --------------------------------------------- We just don't want to connect to the lives nodes for security and quality reasons. As I mention previously we have two clusters. One dedicated to take request from live users(ONLINE DC) and one for all machine learning computation(OFFLINE DC). I prefer the option of specifying the DC since the code avoid any connection to the ONLINE DC. With the CL.ONE if a replica in our OFFLINE DC is down it will use the ONLINE DC and we are trying to avoid that. Thoughts on this front? I'm ok with it if you think that keeping it simple is the best choice in which case I'll have to build from sources and apply my patch. > Ability to specify a DC to consume from when using ColumnFamilyInputFormat > externally > ------------------------------------------------------------------------------------- > > Key: CASSANDRA-6124 > URL: https://issues.apache.org/jira/browse/CASSANDRA-6124 > Project: Cassandra > Issue Type: Improvement > Components: Hadoop > Reporter: Patricio Echague > Priority: Minor > Labels: hadoop > Fix For: 1.2.11 > > Attachments: CASSANDRA-6124.diff > > > Our production environment looks like this: > - 6 cassandra nodes (online DC) > - 3 cassandra nodes (offline DC) > - Hadoop cluster. > we are interested in connecting to the offline DC from hadoop (not colocated > with cassandra offline dc) > I've tested this patch and seems to work with our 1.2.5 deployment. > Kindly review. -- This message was sent by Atlassian JIRA (v6.1#6144)