You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-issues@hadoop.apache.org by "Xiaobing Zhou (JIRA)" <ji...@apache.org> on 2016/03/24 22:54:25 UTC

[jira] [Commented] (HADOOP-12957) Limit the number of outstanding async calls

    [ https://issues.apache.org/jira/browse/HADOOP-12957?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15211015#comment-15211015 ] 

Xiaobing Zhou commented on HADOOP-12957:
----------------------------------------

I posted patch v000 for review. There could be two changes:
1. make asyncCallCounter as a thread local so it makes each thread being subject to the threshold. But there might be a couple of caller threads doing async calls, which exacerbates buffer usage.
2. make asyncCallCounter configurable.

I'd like to know reviewers' comments on these. Thanks. 

> Limit the number of outstanding async calls
> -------------------------------------------
>
>                 Key: HADOOP-12957
>                 URL: https://issues.apache.org/jira/browse/HADOOP-12957
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: ipc
>            Reporter: Xiaobing Zhou
>            Assignee: Xiaobing Zhou
>         Attachments: HADOOP-12957-HADOOP-12909.000.patch
>
>
> In async RPC, if the callers don't read replies fast enough, the buffer storing replies could be used up. This is to propose limiting the number of outstanding async calls to eliminate the issue.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)