hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anastasia Braginsky (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-18010) Connect CellChunkMap to be used for flattening in CompactingMemStore
Date Mon, 29 May 2017 12:04:04 GMT

    [ https://issues.apache.org/jira/browse/HBASE-18010?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16028264#comment-16028264
] 

Anastasia Braginsky commented on HBASE-18010:
---------------------------------------------

May be the previous scenario was not detailed enough, here is the more detailed one:

1. Chunk C was allocated and got ID 5
1.1 After a while, C and all its cells are not in any use any longer
2. C was reclaimed and removed from mapping, but was returned to pool and is still holding
ID 5
3. C was reused from pool and returned to the user once again
3.1 Cell A was allocated on C (new cell)
4. Cell residing on C is asked what is its chunk id and 5 is the answer
5. ChunkCreator is asked to translate chunk ID 5, but returns null, because C's mapping was
removed at Step 2

bq. when the cell is in use for any reason then we should not be returning that back to pool
right?
In time when chunk was reclaimed all its cells were not in use. Chunk was returned to ChunkCreator
legitimately. The problem is that chunk was reused and its mapping was deleted.

> Connect CellChunkMap to be used for flattening in CompactingMemStore
> --------------------------------------------------------------------
>
>                 Key: HBASE-18010
>                 URL: https://issues.apache.org/jira/browse/HBASE-18010
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Anastasia Braginsky
>
> The CellChunkMap helps to create a new type of ImmutableSegment, where the index (CellSet's
delegatee) is going to be CellChunkMap. No big cells or upserted cells are going to be supported
here.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message