[
https://issues.apache.org/jira/browse/HDFS-7034?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14129267#comment-14129267
]
Jing Zhao commented on HDFS-7034:
---------------------------------
Looks like there are several issues causing the current TestStorageMover flaky:
# A block report, which contains only block deletion, currently happens every
(100 * heartbeat_interval) period of time. Thus before we verify files' storage
policy, we should manually trigger the block report from all the datanodes.
# Currently {{testHotWarmColdDirs}} contains a DISK-only DataNode, and the
total number of DN is 6. Then for a hot file, if its initial 3 hot replicas are
all in DN with both DISK and ARCHIVE, finally we cannot make it "pure cold".
# For {{testNoSpaceDisk}}, the creation of the mover.id file may fail because
of the limited DISK space. Maybe we can add a configuration prop to disable the
mover.id file creation only for tests.
# The current delay time for a DN with error is 10s. In our test, since we only
have 6 DN, and the block deletion report may be delayed, a DN can be marked
with error if its original replica was actually removed but itself was still
selected for block replacement. In this scenario, this long time delay may
cause the migration of other blocks to fail. Thus we should use smaller delay
value in our tests.
Will upload a patch with the above fix later.
> Archival Storage: Fix TestBlockPlacement and TestStorageMover
> -------------------------------------------------------------
>
> Key: HDFS-7034
> URL: https://issues.apache.org/jira/browse/HDFS-7034
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: balancer, namenode
> Reporter: Jing Zhao
> Assignee: Jing Zhao
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
