You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ratis.apache.org by "Rajeshbabu Chintaguntla (JIRA)" <ji...@apache.org> on 2018/08/18 00:48:00 UTC

[jira] [Commented] (RATIS-253) Segmented Raft log rolls over segment frequently

    [ https://issues.apache.org/jira/browse/RATIS-253?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16584542#comment-16584542 ] 

Rajeshbabu Chintaguntla commented on RATIS-253:
-----------------------------------------------

[~msingh][~szetszwo] Here is the patch not including the statemachine data size while calculating entry size.

> Segmented Raft log rolls over segment frequently
> ------------------------------------------------
>
>                 Key: RATIS-253
>                 URL: https://issues.apache.org/jira/browse/RATIS-253
>             Project: Ratis
>          Issue Type: Bug
>          Components: server
>    Affects Versions: 0.2.0
>            Reporter: Mukul Kumar Singh
>            Assignee: Rajeshbabu Chintaguntla
>            Priority: Major
>              Labels: ozone
>         Attachments: RATIS-253.patch
>
>
> With statemachine implementation which abstract out the state machine data from the append entries, frequent log rollovers can be observed.
> Looking into the code, this seems to be because of 2 reasons.
> a) LogSegment#append, updates the total size. However the size which needs to be considered should not include the statemachine data.
> b) Also SegmentRaftLog#isSegmentFull also considers the total proto object. Here too, only the entry to be written to the log should be considered.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)