[
https://issues.apache.org/jira/browse/HBASE-16438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15936900#comment-15936900
]
Anoop Sam John commented on HBASE-16438:
----------------------------------------
Not clearing/removing from Q is not the real issue. Code wise there may not be
a remove. But the MSLABImpl object keeps this Q and the MSLAB object become
dead once the segment referring it gets flushed. So some time we will remove
it. The issue was specific to use case where there are many updates to same
Cell. MSLAB was in use in that! So we have cells from one chunk but all
become irrelevant after some time as new cells comes in. Adding new cells to
CSLM would remove the old. If we were not referring to that chunk any where
other than those original Cells, a GC can collect it. But adding to the Q was
preventing it. That is why that jira changed it to keep only chunks from pool.
Now we keep refer to all chunks created out of ChunkCreator. So we are back to
old problem
But as such there wont be OOME for sure. Because as part of above jira
[~carp84] fixed another issue where we account for the cell size (when its data
bytes in chunk) and so we will end up in size breaching at region level/ global
level and normal/forced flush will happen. So I dont feel like any OOME issue.
But to have a better GC in this scenario, HBASE-16195 would have helped ,
which will be broken by this jira.
Am I making things clear?
cc [~carp84]
> Create a cell type so that chunk id is embedded in it
> -----------------------------------------------------
>
> Key: HBASE-16438
> URL: https://issues.apache.org/jira/browse/HBASE-16438
> Project: HBase
> Issue Type: Sub-task
> Affects Versions: 2.0.0
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Attachments: HBASE-16438_1.patch,
> HBASE-16438_3_ChunkCreatorwrappingChunkPool.patch,
> HBASE-16438_4_ChunkCreatorwrappingChunkPool.patch, HBASE-16438.patch,
> MemstoreChunkCell_memstoreChunkCreator_oldversion.patch,
> MemstoreChunkCell_trunk.patch
>
>
> For CellChunkMap we may need a cell such that the chunk out of which it was
> created, the id of the chunk be embedded in it so that when doing flattening
> we can use the chunk id as a meta data. More details will follow once the
> initial tasks are completed.
> Why we need to embed the chunkid in the Cell is described by [~anastas] in
> this remark over in parent issue
> https://issues.apache.org/jira/browse/HBASE-14921?focusedCommentId=15244119&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15244119
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
