[
https://issues.apache.org/jira/browse/HBASE-16438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15607499#comment-15607499
]
ramkrishna.s.vasudevan commented on HBASE-16438:
------------------------------------------------
Thanks for your comments. The case I think of is this.
We do in memory flushes now. So every in memory flush will create a segment.
The cells are located in this segment. When MSLAB is ON and chunk pool is ON we
create a chunk from the pool and that is where the cells are added.
Now this segment is of type CSLM each entry pointing to the chunk created from
the chunk pool.
When I try to create an index out of this I don't add the chunk where the cell
is added. Instead I just add the chunkid (assume when we ask for a chunk from
the pool it gives out an id) and the cell's offset and length.
Why chunk id is important is because again say there is one more in memory
flush and we create one more segment and the old one is pushed to the pipeline.
So we convert this segment from CSLM to chunk based. Now in my index array
since I have another segment added I need to clearly know for which segment
was this chunk created because the cell's offset and length could be same and
there is no unique way to distinguish.
In the CellArrayMap impl there is no such need because just that they create an
array of cells and every index in the array points to the cell referenced
inside the chunks. We don't need any chunkid there. Because it points to actual
cells. But in CellChunkIndex we don't point to actual cells.
Let me know if I am writing it clearly if not can rephrase again. Thanks.
> 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
>
> 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.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
