accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hudson (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-1243) Multiple assignment may occur if tablet server dies during split
Date Wed, 10 Apr 2013 01:25:14 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-1243?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13627350#comment-13627350
] 

Hudson commented on ACCUMULO-1243:
----------------------------------

Integrated in Accumulo-1.5 #71 (See [https://builds.apache.org/job/Accumulo-1.5/71/])
    ACCUMULO-1243 Added apache header (Revision 1466243)
ACCUMULO-1243 Made tablet loading code only load one tablet when recovering a split.  Made
code more strict, it will only load an exact tablet.  Made load code roll back splits that
have started, but did not create a new tablet.  Added some more test. (Revision 1466211)

     Result = SUCCESS
kturner : 
Files : 
* /accumulo/branches/1.5/server/src/test/java/org/apache/accumulo/server/tabletserver/CheckTabletMetadataTest.java

kturner : 
Files : 
* /accumulo/branches/1.5/core/src/main/java/org/apache/accumulo/core/util/MetadataTable.java
* /accumulo/branches/1.5/server/src/main/java/org/apache/accumulo/server/tabletserver/TabletServer.java
* /accumulo/branches/1.5/server/src/main/java/org/apache/accumulo/server/util/MetadataTable.java
* /accumulo/branches/1.5/server/src/test/java/org/apache/accumulo/server/tabletserver/CheckTabletMetadataTest.java
* /accumulo/branches/1.5/test/src/main/java/org/apache/accumulo/test/functional/SplitRecoveryTest.java
* /accumulo/branches/1.5/test/src/main/java/org/apache/accumulo/test/performance/scan/CollectTabletStats.java
* /accumulo/branches/1.5/test/src/test/java/org/apache/accumulo/test/TestAccumulo1235.java
* /accumulo/branches/1.5/test/src/test/java/org/apache/accumulo/test/TestAccumuloSplitRecovery.java

                
> Multiple assignment may occur if tablet server dies during split
> ----------------------------------------------------------------
>
>                 Key: ACCUMULO-1243
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-1243
>             Project: Accumulo
>          Issue Type: Bug
>    Affects Versions: 1.4.0
>            Reporter: Keith Turner
>            Assignee: Keith Turner
>            Priority: Critical
>             Fix For: 1.5.0, 1.4.4
>
>
> Make the following change to the tablet server code.  The tablet server has to die at
this exact point for the bug to occur.
> {noformat}
> Index: src/main/java/org/apache/accumulo/server/tabletserver/Tablet.java
> ===================================================================
> --- src/main/java/org/apache/accumulo/server/tabletserver/Tablet.java	(revision 1464780)
> +++ src/main/java/org/apache/accumulo/server/tabletserver/Tablet.java	(working copy)
> @@ -3592,6 +3592,9 @@
>        MetadataTable.splitTablet(high, extent.getPrevEndRow(), splitRatio, SecurityConstants.getSystemCredentials(),
tabletServer.getLock());
>        MetadataTable.addNewTablet(low, lowDirectory, tabletServer.getTabletSession(),
lowDatafileSizes, bulkLoadedFiles,
>            SecurityConstants.getSystemCredentials(), time, lastFlushID, lastCompactID,
tabletServer.getLock());
> +      
> +      Runtime.getRuntime().halt(2);
> +
>        MetadataTable.finishSplit(high, highDatafileSizes, highDatafilesToRemove, SecurityConstants.getSystemCredentials(),
tabletServer.getLock());
>        
>        log.log(TLevel.TABLET_HIST, extent + " split " + low + " " + high);
> {noformat}
> Then create a table and add a split.  
> {noformat}
> root@test15> createtable foo
> root@test15 foo> addsplits -t foo m
> {noformat}
> If there are multiple tablet servers, then its possible that multiple assignment may
occur.  Below is an example of this occurring after tablets were loaded.
> {noformat}
> root@test15 !METADATA> scan -b 1 -c loc
> 1;m loc:13d5a86463f4f98 []    127.0.0.1:9998
> 1;m loc:13d5a86463f4f9f []    127.0.0.1:10000
> 1< loc:13d5a86463f4f98 []    127.0.0.1:9998
> {noformat}
> The problem is that the assignment code in the tserver detects an incomplete split and
load both children.   However, the master may also assign one of the children.   
> I think the assignment code should be modified to fix up the metadata table and only
load one tablet. If the new tablet was not created, it should roll back the changes and load
the pre split tablets.  If the new tablet was created, then assume the master will assign
it and only load the high tablet.  I think these changes would greatly simplify the code also.
> I do not think the proposed changes would cause issues with merge, since the chop flag
is deleted in the case where this occurs.
> Need to ensure that the solution is itself fault tolerant.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message