You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "stack (JIRA)" <ji...@apache.org> on 2015/04/02 02:31:53 UTC
[jira] [Commented] (HBASE-12600) Remove REPLAY tag dependency in
Distributed Replay Mode
[ https://issues.apache.org/jira/browse/HBASE-12600?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14391859#comment-14391859 ]
stack commented on HBASE-12600:
-------------------------------
[~jeffreyz] One side-effect is that we preserve the seqid across compactions -- which makes sense I suppose (we don't want our DLR seqid going away if compaction while replay) -- but downside is that it undoes the optimization made in HBASE-8151 (compactions let go of seqids if no scanner outstanding has a readpoint behind oldest edit in storefiles).
An edit could come in with a seqid before that of an edit in a storefile when replaying so we need this on for correctness-sake? What you reckon?
> Remove REPLAY tag dependency in Distributed Replay Mode
> -------------------------------------------------------
>
> Key: HBASE-12600
> URL: https://issues.apache.org/jira/browse/HBASE-12600
> Project: HBase
> Issue Type: Bug
> Components: wal
> Affects Versions: 2.0.0, 0.99.1
> Reporter: Jeffrey Zhong
> Assignee: Jeffrey Zhong
> Fix For: 2.0.0, 0.99.2
>
> Attachments: HBASE-12600.patch
>
>
> After HBASE-11315 & HBASE-8763, each edit has a unique 'version' i.e. its SequenceId(or old mvcc value). Therefore, we don't need replay tag to handle out of order same version updates.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)