hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Varun Sharma (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-9503) Allow a configurable sleep between IPC client connect timeouts
Date Thu, 25 Apr 2013 00:19:13 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-9503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13641214#comment-13641214
] 

Varun Sharma commented on HADOOP-9503:
--------------------------------------

[~szetszwo]
Hi Nicholas,

This configures things when you have IOException but for SocketTimeout(s) we had a hard coded
1s sleep with a hard coded 20second connect timeout and hard coded 45 retries. Both were fixed
but still there is a constant sleep of 1 second b/w retries for socket timeout errors (which
you can get when a host black holed - like firewalled to simulate network disasters).
                
> Allow a configurable sleep between IPC client connect timeouts
> --------------------------------------------------------------
>
>                 Key: HADOOP-9503
>                 URL: https://issues.apache.org/jira/browse/HADOOP-9503
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: ipc
>            Reporter: Varun Sharma
>            Priority: Minor
>
> HADOOP 9106 introduced a configurable retry timeout when there are socket timeouts errors.
> Since the IPC client is used by a wide variety of applications including realtime/online
ones (like hbase + hdfs), we sometimes need fast fail timeouts which are < 3 seconds. Achieving
such a timeout for connect is very difficult with a 1 second sleep in b/w retries. It would
be good to have a configurable parameter to reduce this sleep. 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message