[
https://issues.apache.org/jira/browse/MAPREDUCE-7476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17843152#comment-17843152
]
ASF GitHub Bot commented on MAPREDUCE-7476:
-------------------------------------------
kaiyaok2 opened a new pull request, #6790:
URL: https://github.com/apache/hadoop/pull/6790
### Description of PR
Similar with https://issues.apache.org/jira/browse/MAPREDUCE-7475 , this PR
fixes more non-idempotent unit tests detected.
## Overview & Proposed Fix of all remaining non-idempotent unit tests in the
MapReduce Project
The following two tests below do not reset `NotificationServlet.counter`, so
repeated runs throw assertion failures due to accumulation.
- org.apache.hadoop.mapred.TestClusterMRNotification#testMR
- org.apache.hadoop.mapred.TestLocalMRNotification#testMR
Fixed by resetting `NotificationServlet.counter` and
`NotificationServlet.failureCounter` to 0 after test execution.
-----------------------------------------------------------------------------------------------
The following test does not remove the key `AMParams.ATTEMPT_STATE`, so
repeated runs of the test will not be missing the attempt-state at all:
- org.apache.hadoop.mapreduce.v2.app.webapp.TestAppController.testAttempts
Fixed by removing `AMParams.ATTEMPT_STATE` at the end of the test.
-----------------------------------------------------------------------------------------------
The following test fully deletes `TEST_ROOT_DIR` after execution, so
repeated runs will throw a`DiskErrorException`:
- org.apache.hadoop.mapred.TestMapTask#testShufflePermissions
Fixed by checking if `TEST_ROOT_DIR` exists before test execution. Make the
directory if not.
-----------------------------------------------------------------------------------------------
The following test does not restore the static variable `statusUpdateTimes`
after execution, so consecutive runs throws `AssertionError`:
- org.apache.hadoop.mapred.TestTaskProgressReporter#testTaskProgress
Fixed by resetting `statusUpdateTimes` to 0 before test execution
### How was this patch tested?
After the patch, rerunning the tests in the same JVM does not produce any
exceptions.
> Follow up of https://issues.apache.org/jira/browse/MAPREDUCE-7475 - detected
> 5 more non-idempotent tests(pass in the first run but fails in repeated runs
> in the same JVM)
> --------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: MAPREDUCE-7476
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-7476
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Kaiyao Ke
> Priority: Major
> Original Estimate: 1h
> Remaining Estimate: 1h
>
> Similar with https://issues.apache.org/jira/browse/MAPREDUCE-7475 , 5 more
> non-idempotent unit tests are detected.
> The following two tests below do not reset `NotificationServlet.counter`, so
> repeated runs throw assertion failures due to accumulation:
> * org.apache.hadoop.mapred.TestClusterMRNotification#testMR
> * org.apache.hadoop.mapred.TestLocalMRNotification#testMR
> The following test does not remove the key `AMParams.ATTEMPT_STATE`, so
> repeated runs of the test will not be missing the attempt-state at all:
> * org.apache.hadoop.mapreduce.v2.app.webapp.TestAppController.testAttempts
> The following test fully deletes `TEST_ROOT_DIR` after execution, so repeated
> runs will throw a`DiskErrorException`:
> * org.apache.hadoop.mapred.TestMapTask#testShufflePermissions
> The following test does not restore the static variable `statusUpdateTimes`
> after execution, so consecutive runs throws `AssertionError`:
> * org.apache.hadoop.mapred.TestTaskProgressReporter#testTaskProgress
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
