You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Jonathan Ellis (JIRA)" <ji...@apache.org> on 2013/11/20 04:19:21 UTC
[jira] [Resolved] (CASSANDRA-5974) Symlink Keys
[ https://issues.apache.org/jira/browse/CASSANDRA-5974?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Jonathan Ellis resolved CASSANDRA-5974.
---------------------------------------
Resolution: Won't Fix
You could do this with triggers. It's awfully specialized for us to add to core.
> Symlink Keys
> ------------
>
> Key: CASSANDRA-5974
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5974
> Project: Cassandra
> Issue Type: New Feature
> Reporter: Jason Fager
> Priority: Minor
>
> Floated this on irc, was told to open a ticket here:
> We have a use case where a 3rd party sends us their id, we do a read to look up our internal id, and then turn around and do a write against that internal id. We'd save a round trip if we could just do the write to our internal id via the external id; i.e., if we could treat the external id as a sort of symlink for our internal id.
> This use case doesn't seem like a great fit for secondary indexing, as there are potentially many external ids for each internal id, but only ever 1 internal id for each external id. Another suggestion on irc was to use triggers; I haven't sat down to figure out if/how that would work, but I think the restriction against counter columns makes a trigger-based approach less than ideal.
> As a vague sketch of how I'd see this working, a symlink row would be distributed just like any other row, but would forward any operations on it to the real row. If the symlink rowkey lands on a different node than the real rowkey, there's an extra network hop, but this is still an improvement over the current case (client > lookup_node > client -> real_node vs client -> symlink_node -> real_node).
--
This message was sent by Atlassian JIRA
(v6.1#6144)