[
https://issues.apache.org/jira/browse/HBASE-18307?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16079840#comment-16079840
]
Duo Zhang commented on HBASE-18307:
-----------------------------------
{quote}
The default thread number is 0?
{quote}
'0' means use the default setting of netty, usually 2 * CPU core count.
{quote}
Why acceptor and client use one EventLoopGroup?
{quote}
See here
https://stackoverflow.com/questions/28331809/netty-bootstrap-with-boss-group-or-with-just-with-workers-eventloopgroup
Norman Maurer is one of the authors of netty. It is OK to use the same event
loop group. The name 'boss' and 'worker' are inherited from netty3. There are
two types of event loop group in netty3 for accept and read/write, and in netty
4 they are unified.
{quote}
How about add a helper class for passing config to NettyRpcServer, similar with
NettyRpcClientConfigHelper and NettyAsyncFSWALConfigHelper? Then we can remove
the cast code in NettyRpcServer.
{quote}
I think cast is simpler and shorter than introduce a new class... And I think
we will finally add a method to the interface, after the cleanup of
IA.LimitedPrivate interfaces in HBASE-18169.
Thanks.
> Share the same EventLoopGroup for NettyRpcServer, NettyRpcClient and
> AsyncFSWALProvider at RS side
> --------------------------------------------------------------------------------------------------
>
> Key: HBASE-18307
> URL: https://issues.apache.org/jira/browse/HBASE-18307
> Project: HBase
> Issue Type: Improvement
> Components: io, rpc, wal
> Affects Versions: 3.0.0, 2.0.0-alpha-1
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Fix For: 3.0.0, 2.0.0-alpha-2
>
> Attachments: HBASE-18307.patch, HBASE-18307-v1.patch
>
>
> Need a find a proper way to pass a EventLoopGroup instance through a
> Configuration object.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)