[
https://issues.apache.org/jira/browse/HBASE-19505?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16289608#comment-16289608
]
Anoop Sam John commented on HBASE-19505:
----------------------------------------
Ya I thought 1st by passing boolean on constructor, we can know RPC server at
HM or RS. I just saw in that RpcServer create path, some place we have an
instance of check based way so made my mind to go that simple way. Ya I can
change it to be boolean based.
On the intelligent way of deciding based on HM carrying region. I think when
HM carry META region, still we might NOT need this BB pool.. Do we expect large
sized mutations write call to Meta? The reads also it is random gets for each
of the region location/state etc. Dont think the read will have to respond
with big sized result. So BBpool might look a waste there IMO. It all depends
on when HM having regions, what regions. Only META (and NS also) or any.
Even the default calc for the #BBs in pool, that depends on the handlers count.
When it is META alone, it will be using the high priority handler and that
count will be relevant. So doing this intelligent full proof way it is not
that easy. And even if HM is having META and NS regions, IMO better not to
have the BBPool. WDYT?
> Disable ByteBufferPool by default at HM
> ---------------------------------------
>
> Key: HBASE-19505
> URL: https://issues.apache.org/jira/browse/HBASE-19505
> Project: HBase
> Issue Type: Sub-task
> Reporter: Anoop Sam John
> Assignee: Anoop Sam John
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-19505.patch
>
>
> The main usage of the pool is while accepting bigger sized requests ie.
> Mutation requests. HM do not have any regions by default. So we can make
> this pool OFF in HM side. Still add a config to turn this ON.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
