phoenix-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (PHOENIX-3472) Mutable index adds DeleteFamily markers which impacts query performance
Date Thu, 10 Nov 2016 18:44:59 GMT

    [ https://issues.apache.org/jira/browse/PHOENIX-3472?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15654797#comment-15654797
] 

James Taylor commented on PHOENIX-3472:
---------------------------------------

We have to issue that delete marker as that's what is deleting the row that represents the
old value to properly maintain the index. We can't know if the old value actually exists or
not (it's potentially a different row on a different server). For immutable tables we don't
because there should never be an old value since a row is never updated (by definition).

It's maybe possible that for local indexes we could check if the index row exists before issuing
the Delete marker since we know it's a local operation. Any thoughts on this [~rajeshbabu]?

> Mutable index adds DeleteFamily markers which impacts query performance
> -----------------------------------------------------------------------
>
>                 Key: PHOENIX-3472
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3472
>             Project: Phoenix
>          Issue Type: Bug
>    Affects Versions: 4.9.0
>            Reporter: Mujtaba Chohan
>
> {noformat}
> create table M (k varchar not null primary key, v archer);
> create index MI on M(v);
> upsert into M values ('a','b');
> {noformat}
> Raw scan shows the following two kv pair with the first one being a delete marker
> {noformat}
> ROW                                     COLUMN+CELL                                 
                                                                     
>  \x00a    column=0:, timestamp=1478801841429, type=DeleteFamily                     
                                       
>  b\x00a   column=0:_0, timestamp=1478801841429, value=_0  
> {noformat}
> This severely impacts read query performance as data size grows, the number of DeleteFamily
markers also grows linearly in the table.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message