hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Colin Patrick McCabe (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-8026) Trace FSOutputSummer#writeChecksumChunks rather than DFSOutputStream#writeChunk
Date Wed, 01 Apr 2015 00:34:53 GMT

    [ https://issues.apache.org/jira/browse/HDFS-8026?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14389768#comment-14389768

Colin Patrick McCabe commented on HDFS-8026:

I think the previous test timeout on {{TestTracing}} was not related to this patch, since
it didn't reproduce the second time I ran it on JIRA.  I also can't reproduce it locally.
 This patch is not really changing how {{TestTracing}} works, just tweaking a string that
it's looking for, so I would not expect it to regress this.  While it's kind of unfortunate
that the test (may be?) flaky, I think we should not let that block this patch.

I'm going to commit in just a bit... if we hit more timeouts on {{TestTracing}}, we'll have
to have a separate JIRA to investigate.  Thanks, [~hitliuyi].

> Trace FSOutputSummer#writeChecksumChunks rather than DFSOutputStream#writeChunk
> -------------------------------------------------------------------------------
>                 Key: HDFS-8026
>                 URL: https://issues.apache.org/jira/browse/HDFS-8026
>             Project: Hadoop HDFS
>          Issue Type: Bug
>    Affects Versions: 2.7.0
>            Reporter: Colin Patrick McCabe
>            Assignee: Colin Patrick McCabe
>            Priority: Minor
>         Attachments: HDFS-8026.001.patch
> We should trace FSOutputSummer#writeChecksumChunks rather than DFSOutputStream#writeChunk.
 When tracing writeChunk, we get a new trace span every 512 bytes; when tracing writeChecksumChunks,
we normally get a new trace span only when the FSOutputSummer buffer is full (9x less often.)

This message was sent by Atlassian JIRA

View raw message