hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "HariSree (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-2017) A partial rollback cause the new changes done after upgrade to be visible after rollback
Date Thu, 02 Jun 2011 08:42:47 GMT

    [ https://issues.apache.org/jira/browse/HDFS-2017?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13042650#comment-13042650

HariSree commented on HDFS-2017:

>You mean to say here the upgrade failed but Namenode started functioning?

Here , it is the Namenode process stopping abnormally(due to external cause) just after upgrade
of the 1st name directory . After that , we're again starting it normally (REGULAR) and it
is starting up fine .. 

>How is this possible? The directory that is rolled back will be consistent with the directories
that were not upgraded previously and hence are not rolled back.

The difference in the directories is due to the second normal restart . Here(the 2nd restart-normal)
, during loadfsimage , the 1st dir ll be loaded and written to other dirs during savenamespace
(chktime incremented) .. Now if new changes are made and  we rollback .. 
 After rollback , since only the 1st dir ll be rollbacked and others left behind (since they
dont 've previous dirs) , the rollbacked 1st dir ll not contain the new changes , while the
others ll contain Delta  .. now the 2nd dir ll be loaded since its chktime is higher than
the 1st dir's chktime , and will be written to all other dirs .. So , due to this , Delta
will be available after this rollback ..

> During rollback new changes are indeed lost.

Yes , thats normal .. 

> A partial rollback cause the new changes done after upgrade to be visible after rollback
> ----------------------------------------------------------------------------------------
>                 Key: HDFS-2017
>                 URL: https://issues.apache.org/jira/browse/HDFS-2017
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: name-node
>    Affects Versions: 0.20.1
>            Reporter: HariSree
>            Priority: Minor
>              Labels: rollback, upgrade
> This is the scenario :
> Namenode has 3 name dirs configured ..
> 1) Namenode upgrade starts - Upgrade fails after 1st directory is upgraded (2nd and 3rd
dir is left unchanged ..) { like , Namenode process down }
> 2) Namenode starts and new files written .. 
> 3) Namenode shutdown and rollbacked
> Since Namenode is saving the latest image dir(the upgraded 1st dir since checkpointtime
is incremented during upgrade for this dir) will be loaded and saved to all dirs during loadfsimage
> But if a ROLLBACK is done , the 1st dir will be rolled back (the older copy becomes current
and its checkpointtime is now LESS than other dirs ..) and others left behind since they dont
contain previous .. Now during loadfsimage , the 2nd dir will be selected since it has the
highest checkpoint time and saved to all dirs (including 1st ) .. Now due to this , the new
changes b/w UPGRADE and ROLLBACK present in 2nd dir gets reflected even after ROLLBACK ..
> This is not the case with a SUCCESSFUL Upgrade/Rollback (New changes lost after rollback)..

This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

View raw message