[
https://issues.apache.org/jira/browse/HDFS-9543?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15258637#comment-15258637
]
Arpit Agarwal commented on HDFS-9543:
-------------------------------------
Hi Anu, thanks for the updated patch. My comments:
# In copyBlocks, the openPoolIters call should be inside the while loop, just
before you open the try block.
# I didn't understand the new computeDelay calculation. Your original
calculation (v1 patch) looked correct, all that remained was to subtract
{{timeUsed}} and check for negative result.
# Thanks for the pointer to {{DirectoryScanner#scan}}. The synchronization is
needed only when making changes to the in-memory block map.
{{FsDatasetImpl#moveBlock}} and its callees already synchronize on the dataset
when updating the map so we should remove the {{synchronized}} block in
{{DiskBalancerMover#copyBlocks}}.
# Nitpick: Line 818: We can also log the maximum error count value here for
quick reference. Also a typo _cound_ -> _count_.
# bq. I will make this comment clearer in the next update. Balancer supports a
flag called -blockpools – HDFS-8890 seems to have added this. Just leaving a
note that we don't do this yet.
I think the -blockpools flag just restricts the balancing to a subset of
blockpools. We cannot copy blocks across blockpools. Would you consider
removing the TODO and filing a Jira if you think a similar flag will be useful
for disk balancer.
Sorry about not catching some of these on the last pass.
> DiskBalancer : Add Data mover
> ------------------------------
>
> Key: HDFS-9543
> URL: https://issues.apache.org/jira/browse/HDFS-9543
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: datanode
> Reporter: Anu Engineer
> Assignee: Anu Engineer
> Attachments: HDFS-9543-HDFS-1312.001.patch,
> HDFS-9543-HDFS-1312.002.patch
>
>
> This patch adds the actual mover logic to the datanode.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
