You are viewing a plain text version of this content. The canonical link for it is here.

Posted to jira@kafka.apache.org by "Matthias J. Sax (Jira)" <ji...@apache.org> on 2020/05/27 20:51:00 UTC

[jira] [Resolved] (KAFKA-40) Time based log rolling

     [ https://issues.apache.org/jira/browse/KAFKA-40?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Matthias J. Sax resolved KAFKA-40.
----------------------------------
    Resolution: Implemented  (was: Abandoned)

It seem the time-index as introduced via KIP-33 ([https://cwiki.apache.org/confluence/display/KAFKA/KIP-33+-+Add+a+time+based+log+index]) addressed this issue.

> Time based log rolling
> ----------------------
>
>                 Key: KAFKA-40
>                 URL: https://issues.apache.org/jira/browse/KAFKA-40
>             Project: Kafka
>          Issue Type: New Feature
>            Assignee: Chris Burroughs
>            Priority: Minor
>
> In some cases we know that consumers are interested in data on time boundaries (for example, hourly), and when a new consumer of the type is spun up it wants to consume data since the last boundary (start at 12 noon). 
> OffsetRequest can do this now, but Log:getOffsetsBefore is "very approximate" and it would be nice for the consumers to not have to iterate over unneeded data that (being older) is less likely to be in the page cache. 
> Proposal: Optional argument to roll log file if it contains more than n seconds of data. I think this is reasonable, but wanted to create a ticket for comments in-case I've missed a reason this infeasible or otherwise a bad idea.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)