[
https://issues.apache.org/jira/browse/HBASE-17343?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15984635#comment-15984635
]
Anoop Sam John edited comment on HBASE-17343 at 4/26/17 11:50 AM:
------------------------------------------------------------------
What I think is as below.. (Did not test and investigated thoroughly though.
I have to do that.. Just as a test way did)
We have a global memstore size above which we dont allow sum of all memstore's
size to grow.. Once we reach we block the writes and make forced flushes. Even
at a lower barrier breach, we take action.
Here while we change the CSLM to CellArray, we have a size increase (temp) and
we dont account that any where.. We were already nearing to the global
memstore limit but and this change would have breached it , blocking further
writes and making forced flushes. (If we accounted that).. The IHOP is so
adjusted considering the global memstore size been 60%. May be like the heap
usage by memstores grow (this is memory which we can not release) above 60% ..
The change might be giving a breathing space as we do some forced flushes.
Again just theory.. Need to do further investigation.
Again this is an issue when we deal with low cell size and higher memstore size
(MSLAB and it is pool).
The change causing the PE avg run time to some more secs. This is as expected.
I tested it as a boundary check so that at any cost, we should not make a RS
crash.
was (Author: anoop.hbase):
What I think is as below.. (Did not test and investigated thoroughly though.
I have to do that.. Just as a test way did)
We have a global memstore size above which we dont allow sum of all memstore's
size to grow.. Once we reach we block the writes and make forced flushes. Even
at a lower barrier breach, we take action.
Here while we change the CSLM to CellArray, we have a size increase (temp) and
we dont account that any where.. We were already nearing to the global
memstore limit but and this change would have breached it , blocking further
writes and making forced flushes. (If we accounted that).. The IHOP is so
adjusted considering the global memstore size been 60%. May be like the heap
usage by memstores grow (this is memory which we can not release) above 60% ..
The change might be giving a breathing space as we do some forced flushes.
Again just theory.. Need to do further investigation.
Again this is an issue when we deal with low cell size and higher memstore size
(MSLAB and it is pool).
> Make Compacting Memstore default in 2.0 with BASIC as the default type
> ----------------------------------------------------------------------
>
> Key: HBASE-17343
> URL: https://issues.apache.org/jira/browse/HBASE-17343
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.0.0
> Reporter: ramkrishna.s.vasudevan
> Priority: Blocker
> Fix For: 2.0.0
>
> Attachments: HBASE-17343-V01.patch, HBASE-17343-V02.patch,
> HBASE-17343-V04.patch, HBASE-17343-V05.patch
>
>
> FYI [~anastas], [~eshcar] and [~ebortnik].
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
