accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Christopher Tubbs (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-1528) Scans should deterministically return entries with identical timestamps
Date Fri, 21 Jun 2013 19:13:19 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-1528?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13690609#comment-13690609
] 

Christopher Tubbs commented on ACCUMULO-1528:
---------------------------------------------

This issue applies to situations where the VersioningIterator is not being used also. Sorry
for the bad summary and description.

Currently, the VersioningIterator will return at most N versions of keys which are unique
in {R,CF,CQ,CV}, but it is not deterministic which ones it will return when there are keys
that are unique including timestamp, {R,CF,CQ,CV,T}.

Currently, without the VersioningIterator, the scan results are still non-deterministic, because
the client can never be certain it has received all versions of the last key in any returned
batch (because there could be more with the exact same timestamp).

Example:

{code}
R1 CF:CQ CV T=1 Value1
R2 CF:CQ CV T=2 Value2
R2 CF:CQ CV T=2 Value3
R2 CF:CQ CV T=2 Value4
R2 CF:CQ CV T=2 Value5
R3 CF:CQ CV T=1 Value6
{code}

Let's say the batch ended with 'R2 CF:CQ CV T=2 Value3'. The client will continue at R3, and
will never see the remaining R2 entries. Additionally, the R2 entries will be returned in
a non-deterministic order, if they come from multiple sources.
                
> Scans should deterministically return entries with identical timestamps
> -----------------------------------------------------------------------
>
>                 Key: ACCUMULO-1528
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-1528
>             Project: Accumulo
>          Issue Type: Sub-task
>          Components: tserver
>            Reporter: Christopher Tubbs
>            Priority: Minor
>
> Scans will return multiple versions of the same key (down to identical timestamps, but
possibly with different values), non-deterministically. A source identity (eg. filename/timestamp)
could be used to order these consistently.

--
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