You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Rajini Sivaram (Jira)" <ji...@apache.org> on 2022/12/08 15:59:00 UTC

[jira] [Resolved] (KAFKA-14352) Support rack-aware partition assignment for Kafka consumers

     [ https://issues.apache.org/jira/browse/KAFKA-14352?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rajini Sivaram resolved KAFKA-14352.
------------------------------------
      Reviewer: David Jacot
    Resolution: Fixed

This includes protocol changes for rack-aware assignment. Default assignors will be made rack-aware under follow-on tickets in the next release.

> Support rack-aware partition assignment for Kafka consumers
> -----------------------------------------------------------
>
>                 Key: KAFKA-14352
>                 URL: https://issues.apache.org/jira/browse/KAFKA-14352
>             Project: Kafka
>          Issue Type: Sub-task
>          Components: consumer
>            Reporter: Rajini Sivaram
>            Assignee: Rajini Sivaram
>            Priority: Major
>             Fix For: 3.4.0
>
>
> KIP-392 added support for consumers to fetch from the replica in their local rack. To benefit from locality, consumers need to be assigned partitions which have a replica in the same rack. This works well when replication factor is the same as the number of racks, since every rack would then have a replica with rack-aware replica assignment. If the number of racks is higher, some racks may not have replicas of some partitions and hence consumers in these racks will have to fetch from another rack. It will be useful to propagate rack in the subscription metadata for consumers and provide a rack-aware partition assignor for consumers.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)