[
https://issues.apache.org/jira/browse/KAFKA-8465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16897788#comment-16897788
]
ASF GitHub Bot commented on KAFKA-8465:
---------------------------------------
lordcheng10 commented on pull request #6866: [KAFKA-8465]replication strategy
for the topic dimension
URL: https://github.com/apache/kafka/pull/6866
When some partiton's replication is assigned to a broker, which disks
should these copies be placed on the broker? The original strategy is to
allocate according to the number of partiitons, but this will cause a partiton
with too many topics to be stored on a disk, which may cause disk hotspot
problems.
In order to solve this problem, we propose an improved strategy: first
ensure that the number of partitions of each disk in the topic dimension is
even. If the number of partitions of a topic on two disks is equal, then sort
according to the total number of partitions on the disk. Select a disk with the
least number of partitions to store the current replication.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Make sure that the copy of the same topic is evenly distributed across a
> broker's disk.
> ---------------------------------------------------------------------------------------
>
> Key: KAFKA-8465
> URL: https://issues.apache.org/jira/browse/KAFKA-8465
> Project: Kafka
> Issue Type: Improvement
> Components: core
> Affects Versions: 2.0.1, 2.2.0, 2.2.1
> Reporter: ChenLin
> Priority: Major
> Attachments: image-2019-07-30-13-40-12-711.png,
> image-2019-07-30-13-40-49-878.png,
> replication_strategy_for_the_topic_dimension.patch
>
>
> When some partiton's replication is assigned to a broker, which disks should
> these copies be placed on the broker? The original strategy is to allocate
> according to the number of partiitons。This strategy will result in uneven
> disk allocation for the topic dimension.
> In order to solve this problem, we propose an improved strategy: first
> ensure that the number of partitions of each disk in the topic dimension is
> even. If the number of partitions of a topic on two disks is equal, then sort
> according to the total number of partitions on the disk. Select a disk with
> the least number of partitions to store the current replication.
> !image-2019-07-30-13-40-12-711.png!
> !image-2019-07-30-13-40-49-878.png!
> Â
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
