[
https://issues.apache.org/jira/browse/HBASE-14978?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15061991#comment-15061991
]
Anoop Sam John commented on HBASE-14978:
----------------------------------------
Ok lets us target first some thing for 1.2 release.
bq.if (!(c instanceof BufferedDataBlockEncoder.SeekerState)) {
So you don't want to do this barrier check for DBE cases? When DBEs we will be
returning ClonedSeekerState as Cells. May be continue with what you were doing
in old patch? (Consider valueArray) So you can even do it for DBE case. For
PrefixTree also value buffer will be the same byte[] which backs the HFileBlock
I believe (Correct Ram?)
Here itself we are seeing the complexity and tomorrow if a new DBE comes in, we
might break the assumption here.. That was my worry.. The code is basically
breaking the Cell notion.. I agree to your point that more accurate house
keeping might be too much of overhead and we need a fix for the OOME
possibility. Just saying my views. :-)
> Don't allow Multi to retain too many blocks
> -------------------------------------------
>
> Key: HBASE-14978
> URL: https://issues.apache.org/jira/browse/HBASE-14978
> Project: HBase
> Issue Type: Improvement
> Affects Versions: 2.0.0, 1.2.0, 1.3.0
> Reporter: Elliott Clark
> Assignee: Elliott Clark
> Priority: Blocker
> Fix For: 1.2.0, 3.0.0, 1.3.0
>
> Attachments: HBASE-14978-branch-1.2.patch, HBASE-14978-v1.patch,
> HBASE-14978-v2.patch, HBASE-14978-v3.patch, HBASE-14978-v4.patch,
> HBASE-14978.patch
>
>
> Scans and Multi's have limits on the total size of cells that can be
> returned. However if those requests are not all pointing at the same blocks
> then the KeyValues can keep alive a lot more data than their size.
> Take the following example:
> A multi with a list of 10000 gets to a fat row. Each column being returned in
> in a different block. Each column is small 32 bytes or so.
> So the total cell size will be 32 * 10000 = ~320kb. However if each block is
> 128k then total retained heap size will be almost 2gigs.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
