[
https://issues.apache.org/jira/browse/HBASE-19930?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16357308#comment-16357308
]
Chia-Ping Tsai commented on HBASE-19930:
----------------------------------------
{quote}When the in memory flush happens to a CCM, then itself all cells
(including any big cells) should get copied to MSLAB
{quote}
When the in memory flush happens, the first ImmutableSegment added to pipeline
is CSLMImmutableSegment. Hence, it has a good chance that a
CSLMImmutableSegment exists in the merge group when the count of segments
reaches the threshold (the action is changed from flatten to merge).
> fix ImmutableMemStoreLAB#forceCopyOfBigCellInto
> -----------------------------------------------
>
> Key: HBASE-19930
> URL: https://issues.apache.org/jira/browse/HBASE-19930
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0-beta-1
> Reporter: Gali Sheffi
> Assignee: Gali Sheffi
> Priority: Major
> Attachments: HBASE-19930-V01.patch, HBASE-19930-V02.patch,
> HBASE-19930-V03.patch, HBASE-19930-V04.patch
>
>
> This issue is about fixing ImmutableMemStoreLAB#forceCopyOfBigCellInto.
> Following a comment in HBASE-19133 regarding a bug in
> ImmutableMemStoreLAB#forceCopyOfBigCellInto (assuming this method is never
> called for an ImmutableMemStoreLAB, and just throwing an
> IllegalStateException whenever called), the forceCopyOfBigCellInto method now
> performs the copy of big cells on the first MSLABImpl in its mslabs
> linked-list.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
