You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@rocketmq.apache.org by "Jaskey Lam (JIRA)" <ji...@apache.org> on 2017/04/12 12:57:41 UTC

[jira] [Comment Edited] (ROCKETMQ-169) Support flow control when slave restarts and replicating from master

    [ https://issues.apache.org/jira/browse/ROCKETMQ-169?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15965784#comment-15965784 ] 

Jaskey Lam edited comment on ROCKETMQ-169 at 4/12/17 12:57 PM:
---------------------------------------------------------------

I think this is an important and beneficial feature, since if a new slave or slave who is falling behind much starts, the out bandwidth will be occupied much and this will influence the consuming tps and synchronization of another healthy slave.

A trafic flow control should be done.

In my opinion, we can add the traffic control code logic in the master when process reading data from slave for:
    {%code%}
    processWriteEvent()...//existing trasfer logic
    ...
    if (isSlaveRequestAVerySmallOffset) {
         computeTpsAndDoFlowControl();//sleep for while 
    } 

   {%code%}

[~Yukon]
[~lizhanhui]
[~vongosling]

what's your advice?


was (Author: jaskey):
I think this is an important and beneficial feature, since if a new slave or slave who is falling behind much starts, the out bandwidth will be occupied much and this will influence the consuming tps and synchronization of another healthy slave.

A trafic flow control should be done.

In my opinion, we can add the traffic control code logic in the master when process reading data from slave for:
    
    processWriteEvent()...
    ...
    if (isSlaveRequestAVerySmallOffset) {
         computeTpsAndDoFlowControl();//sleep for while 
    } 


[~Yukon]
[~lizhanhui]
[~vongosling]

what's your advice?

> Support flow control when slave restarts and replicating from master
> --------------------------------------------------------------------
>
>                 Key: ROCKETMQ-169
>                 URL: https://issues.apache.org/jira/browse/ROCKETMQ-169
>             Project: Apache RocketMQ
>          Issue Type: Improvement
>          Components: rocketmq-broker
>    Affects Versions: 4.0.0-incubating
>            Reporter: Eason Chen
>            Assignee: yukon
>
> as we know slave replicates message from master, but when slave restarts the replicating will cause huge traffic which will impact sending messages, so we suggest supporting flow control for slave when restarts.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)