hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Plamen Jeliazkov (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-7906) BlockReport of a RUR can have a FINALIZED original replica from Truncate
Date Mon, 23 Mar 2015 17:37:11 GMT

    [ https://issues.apache.org/jira/browse/HDFS-7906?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14376240#comment-14376240

Plamen Jeliazkov commented on HDFS-7906:

I can confirm there is no problem.

Ran several times and did a review.
We are not hitting the assertion anymore and the BlockListAsLongs is not sending the block
with a recovery id so the block is added as a FINALIZED block as expected. 

> BlockReport of a RUR can have a FINALIZED original replica from Truncate
> ------------------------------------------------------------------------
>                 Key: HDFS-7906
>                 URL: https://issues.apache.org/jira/browse/HDFS-7906
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: test
>    Affects Versions: 2.7.0
>            Reporter: Plamen Jeliazkov
>            Assignee: Plamen Jeliazkov
> HDFS-7886 revealed that there is a new replica state change as part of Truncate.
> It is now possible that a FINALIZED replica become a RUR replica with an original replica
pointing back to the FINALIZED replica.
> As a result, it is possible to hit an assert in BlockLIstAsLongs.setBlock() whereby you
are trying to add a FINALIZED replica to the ucList because a RUR only had an under-construction
original replica prior to Truncate.

This message was sent by Atlassian JIRA

View raw message