[
https://issues.apache.org/jira/browse/HDFS-5390?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Arpit Agarwal updated HDFS-5390:
--------------------------------
Attachment: h5390.04.patch
{quote}
# Question: In BPServiceActor.reportReceivedDeletedBlocks(), should we
combine the reports and send it once since
DatanodeProtocolClientSideTranslatorPB.blockReceivedAndDeleted(..) could pass
an array of storages? Or we should change blockReceivedAndDeleted(..) to handle
only one storage (we can do it separately)?
{quote}
Good catch, we can probably send all the reports in one message. I would need
to take a look at the error handling, it is easy with a single report. Is it
okay if I file a separate Jira for this?
I addressed all your remaining points.
Thanks.
> Incremental block reports should be sent per storage directory
> --------------------------------------------------------------
>
> Key: HDFS-5390
> URL: https://issues.apache.org/jira/browse/HDFS-5390
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: datanode
> Affects Versions: Heterogeneous Storage (HDFS-2832)
> Reporter: Arpit Agarwal
> Assignee: Arpit Agarwal
> Attachments: h5390.03.patch, h5390.04.patch
>
>
> Incremental block reports should be sent per storage directory since the
> NameNode now expects all block reports to be sent per storage. Currently the
> Datanode tracks pending incremental block reports for all storages in a
> single queue.
> This is a continuation of the work started by HDFS-5134 and HDFS-4988 to
> split block processing per storage directory.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
