[
https://issues.apache.org/jira/browse/HDFS-784?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12798232#action_12798232
]
Suresh Srinivas commented on HDFS-784:
--------------------------------------
Patch in HDFS-127 resets DFSInputStream.failures to zero. This causes infinite
loop when DFSInputStream.read() is called for a block that has all replicas
corrupt. Before this patch, based on failure count reaching a threshold, the
read attempt was abandoned. Since failure count does not reach the threshold
after the change in HDFS-127, DFSClient attempts to read the block in a loop.
The patch from HDFS-127 should be reverted.
> TestFsck times out on branch 0.20.1
> -----------------------------------
>
> Key: HDFS-784
> URL: https://issues.apache.org/jira/browse/HDFS-784
> Project: Hadoop HDFS
> Issue Type: Test
> Affects Versions: 0.20.1
> Reporter: Eli Collins
> Priority: Blocker
> Fix For: 0.20.2
>
>
> Two tests are currently failing on Hudson:
> chttp://hudson.zones.apache.org/hudson/view/Hadoop/job/Hadoop-20-Build/lastBuild/testReport
> >>> org.apache.hadoop.hdfs.TestDatanodeBlockScanner.testBlockCorruptionPolicy
> >>> 0.0020 9
> >>> org.apache.hadoop.hdfs.server.namenode.TestFsck.testCorruptBlock
> HDFS-734 has already been filed for TestDatanodeBlockScanner.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
