[
https://issues.apache.org/jira/browse/KAFKA-3300?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15171293#comment-15171293
]
ASF GitHub Bot commented on KAFKA-3300:
---------------------------------------
GitHub user becketqin opened a pull request:
https://github.com/apache/kafka/pull/983
KAFKA-3300: Avoid over allocating disk space and memory for index files.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/becketqin/kafka KAFKA-3300
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/983.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #983
----
commit b49a9af4c19513e458ced92ef49504f7a1c237df
Author: Jiangjie Qin <[email protected]>
Date: 2016-02-29T01:39:18Z
KAFKA-3300: Avoid over allocating disk space and memory for index files.
----
> Calculate the initial/max size of offset index files and reduce the memory
> footprint for memory mapped index files.
> -------------------------------------------------------------------------------------------------------------------
>
> Key: KAFKA-3300
> URL: https://issues.apache.org/jira/browse/KAFKA-3300
> Project: Kafka
> Issue Type: Improvement
> Affects Versions: 0.9.0.1
> Reporter: Jiangjie Qin
> Assignee: Jiangjie Qin
> Fix For: 0.10.0.0
>
>
> Currently the initial/max size of offset index file is configured by
> {{log.index.max.bytes}}. This will be the offset index file size for active
> log segment until it rolls out.
> Theoretically, we can calculate the upper bound of offset index size using
> the following formula:
> {noformat}
> log.segment.bytes / index.interval.bytes * 8
> {noformat}
> With default setting the bytes needed for an offset index size is 1GB / 4K *
> 8 = 2MB. And the default log.index.max.bytes is 10MB.
> This means we are over-allocating at least 8MB on disk and mapping it to
> memory.
> We can probably do the following:
> 1. When creating a new offset index, calculate the size using the above
> formula,
> 2. If the result in (1) is greater than log.index.max.bytes, we allocate
> log.index.max.bytes instead.
> This should be able to significantly save memory if a broker has a lot of
> partitions on it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
