[ https://issues.apache.org/jira/browse/KAFKA-3062?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ray Chiang updated KAFKA-3062: ------------------------------ Component/s: replication > Read from kafka replication to get data likely Version based > ------------------------------------------------------------ > > Key: KAFKA-3062 > URL: https://issues.apache.org/jira/browse/KAFKA-3062 > Project: Kafka > Issue Type: Improvement > Components: replication > Reporter: xingang > Priority: Major > > Since Kafka require all the reading happen in the leader for the consistency. > in case, for large volume data of a Topic have a number of consumers, among > them, some consumers are Not latency-sensitive But Data-Loss sensitive have a > consuming lag, then, it will pollute the Page cache as they are reading from > the disk, hence, for other consumers which are latency-sensitive will be sad > due to the pagecache miss > If there could be possible for the reading can happen in replicas for a > eventually consistent (likely version based), thus, consumers marked as "read > from replica" can be separated from others -- This message was sent by Atlassian JIRA (v7.6.3#76005)