[
https://issues.apache.org/jira/browse/MAPREDUCE-6519?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14979941#comment-14979941
]
Nemanja Matkovic commented on MAPREDUCE-6519:
---------------------------------------------
TL; DR; I think we can commit this patch as it doesn't introduce regressions
Two QA bot runs are running identical code, only two test cases with same
failure are:
* hadoop.mapreduce.v2.TestMRJobsWithHistoryService
* hadoop.mapred.TestMiniMRClientCluster
Both of these have very wild history full of reds and blues in pass/fail chart.
In order to be good citizen (and sleep better at night) I went back and re-ran
all these failed tests locally on java 7 and java 8 (named java7 and java8
below).
I also ran them all on both java 7 and java 8 without any of IPv6 changes (head
being eb6379ca25e1bb6d3978bd3a021723c38c95bec9). These are named base_java7 and
base_java8 below.
These are local results I have and investigation of history of these test cases
that I saw:
* hadoop.mapreduce.filecache.TestClientDistributedCacheManager
** java7, java8, base_java7, base_java8: FAIL. This is consistently failing
because it fails here:
"Assert.assertFalse(jobConf.getBoolean(MRJobConfig.CACHE_FILE_VISIBILITIES,true));"
and I haven't changed this configuration.
* hadoop.mapreduce.v2.TestMRJobsWithProfiler
** java7, base_java7: Pass
** java8, base_java8: Fail
** Consistent behavior, no regression
** There is no history data for this test case
* hadoop.mapred.TestMiniMRClientCluster
** java7, base_java7: Fail
** java8, base_java8: Pass
** Consistent behavior, not regression
** History data shows lots of flakiness with same failure
* hadoop.mapreduce.v2.TestMRJobs
** java7, java8, base_java7, base_java8: Pass
** Failure in QABot run looks strange, but similar to the way other flaky tests
are failing
* hadoop.mapreduce.v2.TestNonExistentJob
** java7, java8, test_java7, test_java8: PASS
** There is one previous run QA bot with same failure, so very occasional
flakyness
* hadoop.mapreduce.v2.TestUberAM
* java7, java8, base_java7: Pass
* base_java8: Timeout
* No regression
* hadoop.mapreduce.v2.TestMRJobsWithHistoryService
** java7: Multiple runs got both pass and fail locally
** java8: Pass
** base_java7: Fail
** base_java8: Pass
** Test historically flaky
> hadoop-mapreduce - Avoid unsafe split and append on fields that might be IPv6
> literals
> --------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-6519
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-6519
> Project: Hadoop Map/Reduce
> Issue Type: Task
> Reporter: Nemanja Matkovic
> Assignee: Nemanja Matkovic
> Labels: ipv6
> Attachments: MAPREDUCE-6519-HADOOP-11890.1.patch,
> MAPREDUCE-6519-HADOOP-11890.2.patch
>
> Original Estimate: 48h
> Remaining Estimate: 48h
>
> mapreduce portion of patch in HADOOP-12122 that couldn't run due to number of
> components touched.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
