[ 
https://issues.apache.org/jira/browse/HDFS-12711?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16256584#comment-16256584
 ] 

Allen Wittenauer commented on HDFS-12711:
-----------------------------------------

Ignoring the hs_err_pid log files is pretty much just sticking our collective 
heads in the sand about actual, real problems with the unit tests. The unit 
tests themselves haven't been rock solid for a very long time, even before all 
of this start happening.   Entries have been put into the ignore pile so often 
that I wouldn't be surprised if the community is already at the point that most 
developers are ignoring precommit.  (e.g., commits with findbugs reported in 
the issues, javadoc compilation failures being treated as "environmental", etc, 
etc.) 

If I were actually paying more attention to day-to-day Hadoop bits these days, 
I'd probably be ready to disable unit tests (at least HDFS) to specifically 
avoid the "cried wolf" condition.  The rest of the precommit tests work 
properly the vast majority of the time and are probably more important given 
the current state of things. (Never mind the massive speed up. QBT is hitting 
the 15 hour mark for a full run for branch-2 when it is actually allowed to 
complete.)  No one seems to actually care that the unit tests are a broken mess 
and I doubt they'd be missed.

My goal here was to prevent Hadoop from bringing down the rest of the ASF build 
infrastructure.  It's under enough stress without this project making things 
that much worse.  Achievement unlocked and other Yetus users will pick up those 
new safety features in the next release.  I should probably close this JIRA 
issue. Unless someone else plans to spend some effort on these bugs?  At least 
at this point in time, I view my work here as complete. 

Also:

{code}
/build/
{code}

ARGH.  That hasn't been valid since Hadoop used ant.  A great example of "well, 
if we ignore it, it doesn't exist, right?"  Because anything that is still 
using /build/ almost certainly isn't safe for parallel tests and likely 
contributing to a whole host of problems.

> deadly hdfs test
> ----------------
>
>                 Key: HDFS-12711
>                 URL: https://issues.apache.org/jira/browse/HDFS-12711
>             Project: Hadoop HDFS
>          Issue Type: Test
>    Affects Versions: 2.9.0, 2.8.2
>            Reporter: Allen Wittenauer
>            Priority: Critical
>         Attachments: HDFS-12711.branch-2.00.patch, fakepatch.branch-2.txt
>
>




--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to