hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Raghu Angadi (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-4556) Block went missing
Date Mon, 10 Nov 2008 20:09:44 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-4556?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12646345#action_12646345

Raghu Angadi commented on HADOOP-4556:

> 1. removeStoredBlock makes sure that the replica also gets removed from recentInvalidatesSet;
the patch removes it from the other set.

So we have two sets : one in dn and one in FSNamesystem. These two are cleared in two different
places (and two different reasons): dn's is cleared when a datanode is marked dead and FSNamesystem's
is cleared when a block is deleted.

I think for consistency and clarity, both should be cleared in one place in the code.

The patch as is I think fixes the problem.. so if we don't want to change the patch, it might
be ok for 0.18. But for 19 and trunk at least I think it is better to implement this policy
in one place.

> Block went missing
> ------------------
>                 Key: HADOOP-4556
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4556
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: dfs
>    Affects Versions: 0.17.2
>            Reporter: Robert Chansler
>            Assignee: Hairong Kuang
>             Fix For: 0.18.3
>         Attachments: missing-block-example.txt, staleInvalidates.patch, staleInvalidates1.patch
> Suspicion that all replicas of a block were marked for deletion. (Don't panic, investigation

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message