You are viewing a plain text version of this content. The canonical link for it is here.
Posted to hdfs-issues@hadoop.apache.org by "Wei-Chiu Chuang (JIRA)" <ji...@apache.org> on 2018/04/16 01:12:00 UTC
[jira] [Commented] (HDFS-12906) hedged point read in DFSInputStream
sends only 1 hedge read request
[ https://issues.apache.org/jira/browse/HDFS-12906?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16438904#comment-16438904 ]
Wei-Chiu Chuang commented on HDFS-12906:
----------------------------------------
Hi [~thinktaocs] could you please add affect version? We fixed quite a few hedged read bugs late last year (HDFS-11738, HDFS-11303, HDFS-11708) so I'm not sure if this issue is still valid.
> hedged point read in DFSInputStream sends only 1 hedge read request
> -------------------------------------------------------------------
>
> Key: HDFS-12906
> URL: https://issues.apache.org/jira/browse/HDFS-12906
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: hdfs-client
> Reporter: Tao Zhang
> Assignee: Tao Zhang
> Priority: Major
>
> Hedged point read is handled in DFSInputStream.hedgedFetchBlockByteRange(). It calls "getFirstToComplete()" to get the 1st returned result after sending out hedge read requests. But since "getFirstToComplete()" uses "CompletionService.take()" which is a endlessly blocking operation. It will wait for 1 result after sending only 1 hedge read request.
> It could be changed to wait for a specific timeout (instead of infinite timeout) and starting another hedge read request.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org