[
https://issues.apache.org/jira/browse/HDFS-9626?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15088001#comment-15088001
]
Xiao Chen commented on HDFS-9626:
---------------------------------
The failed "Testcase 4" is as follows:
Initially have blocks on rack0, rack1, rack2, and replicate a new block to
{{newNode}} on rack2. Then give a invalid delhint, and wait for the extra
replica to be deleted.
The test expects that, when wrong delhint is given, the deletion always happens
on rack2 (where 2 replicas are located). However, with the change of HDFS-9314,
this is not guaranteed. Instead, the guarantee is after deletion, number of
racks is still >=2. (See [comments in
HDFS-9314|https://issues.apache.org/jira/browse/HDFS-9314?focusedCommentId=15014734&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-15014734]
for details.)
Thus, patch 1 proposes to loose the check in this test, and only to make sure
the extra replica is deleted. The number of racks guarantee is already in place
in {{TestReplicationPolicy}} as part of HDFS-9314.
> TestBlockReplacement#testBlockReplacement fails occasionally
> ------------------------------------------------------------
>
> Key: HDFS-9626
> URL: https://issues.apache.org/jira/browse/HDFS-9626
> Project: Hadoop HDFS
> Issue Type: Test
> Reporter: Xiao Chen
> Assignee: Xiao Chen
> Attachments: HDFS-9626.01.patch
>
>
> testBlockPlacement sometimes fail in test case 4 in {{checkBlocks}}. I'll
> post the detailed error in comment.
> Thanks [~jojochuang] for helping identify the issue.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
