[
https://issues.apache.org/jira/browse/HDFS-9138?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14906861#comment-14906861
]
Chris Nauroth commented on HDFS-9138:
-------------------------------------
Similar to HDFS-4732, this can be fixed by avoiding hard links in the
checked-in tarball and instead using copies of the files.
> TestDatanodeStartupFixesLegacyStorageIDs fails on Windows due to failure to
> unpack old image tarball that contains hard links
> -----------------------------------------------------------------------------------------------------------------------------
>
> Key: HDFS-9138
> URL: https://issues.apache.org/jira/browse/HDFS-9138
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: test
> Reporter: Chris Nauroth
> Assignee: Chris Nauroth
> Priority: Minor
>
> {{TestDatanodeStartupFixesLegacyStorageIDs#testUpgradeFrom22via26FixesStorageIDs}}
> uses a checked-in DataNode data directory that contains hard links. The
> hard links cannot be handled correctly by the commons-compress library used
> in the Windows implementation of {{FileUtil#unTar}}. The result is that the
> unpacked block files have 0 length, the block files reported to the NameNode
> are invalid, and therefore the mini-cluster never gets enough good blocks
> reported to leave safe mode.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
