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 "Hanisha Koneru (JIRA)" <ji...@apache.org> on 2017/09/22 21:03:01 UTC
[jira] [Commented] (HADOOP-14902) LoadGenerator#genFile write close
timing is incorrectly calculated
[ https://issues.apache.org/jira/browse/HADOOP-14902?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16177115#comment-16177115 ]
Hanisha Koneru commented on HADOOP-14902:
-----------------------------------------
Thanks for reporting this bug, [~jlowe].
As per your [comment|https://issues.apache.org/jira/browse/HADOOP-14881?focusedCommentId=16177074&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16177074] in HADOOP-14881, I agree that we should not do double close as a norm.
And I think we should not update the metric if close fails. The metric tracks the execution time for close operation. If we add the time when exception is thrown, it would pollute the metric.
What do you think?
> LoadGenerator#genFile write close timing is incorrectly calculated
> ------------------------------------------------------------------
>
> Key: HADOOP-14902
> URL: https://issues.apache.org/jira/browse/HADOOP-14902
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs
> Affects Versions: 2.4.0
> Reporter: Jason Lowe
> Assignee: Hanisha Koneru
>
> LoadGenerator#genFile's write close timing code looks like the following:
> {code}
> startTime = Time.now();
> executionTime[WRITE_CLOSE] += (Time.now() - startTime);
> {code}
> That code will generate a zero (or near zero) write close timing since it isn't actually closing the file in-between timestamp lookups.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org