Create separate read repair settings for intra- versus inter- datacenter
------------------------------------------------------------------------
Key: CASSANDRA-2970
URL: https://issues.apache.org/jira/browse/CASSANDRA-2970
Project: Cassandra
Issue Type: Improvement
Components: Core
Reporter: Jeremy Hanna
Priority: Minor
When doing read repair, it doesn't take into account the datacenter where the
replicas are. It simply does a repair based on the read repair chance. Since
multi-DC configurations would benefit from a lower chance between DC, it seems
reasonable to have a separate setting for read repair between DCs than for
within the DC.
Perhaps there could be a single property still, which would default both inter
and intra for the sake of simple scenarios and backwards compatibility. Then
if a more specific setting is specified (read_repair_chance_global or
read_repair_chance_local), that would be used.
I'm not sure if this would complicate matters too much, but it builds on
CASSANDRA-982 and CASSANDRA-1530 to help with efficiency between datacenters,
especially in the case of an analytics cluster versus a realtime cluster -
making them tunably more independent.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
