hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ramkrishna.s.vasudevan (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-16500) Do not add any store file on scanner reset that has a seq id greater than current read pt
Date Thu, 25 Aug 2016 10:54:21 GMT

    [ https://issues.apache.org/jira/browse/HBASE-16500?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15436660#comment-15436660
] 

ramkrishna.s.vasudevan commented on HBASE-16500:
------------------------------------------------

The simplest way to solve this in in FileInfo record the lowest possible memstoreTS. So we
have maxMemstoreTS in StoreFileREader now. Similarly we will have minMemstoreTS in that file.
Compare the readPt with this minMemstoreTS and if the readPt is greater then don't include
that file. 

> Do not add any store file on scanner reset that has a seq id greater than current read
pt
> -----------------------------------------------------------------------------------------
>
>                 Key: HBASE-16500
>                 URL: https://issues.apache.org/jira/browse/HBASE-16500
>             Project: HBase
>          Issue Type: Improvement
>          Components: Performance, Scanners
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>             Fix For: 2.0.0
>
>
> This in principle is similar to HBASE-15871. 
> Take for eg. we have a scanner created with read pt 4. After which a we do a lot of mutations
and then flush the region. 
> Now when the scanner reset happens as part of the scan flow we tend to add the the newly
flushed file also to the scanner heap and we keep scanning the file only to see that all the
KVs are having an mvcc greater than the read pt 4. Ideally it is better to skip all of them
and not add in the scanner heap. 



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

Mime
View raw message