[
https://issues.apache.org/jira/browse/HDFS-12904?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16917514#comment-16917514
]
Lisheng Sun commented on HDFS-12904:
------------------------------------
[~elgoiri] DataXceiver#writeBlock
{code:java}
blockReceiver.receiveBlock(mirrorOut, mirrorIn, replyOut,
mirrorAddr, null, targets, false);
}
{code}
As above code, DataXceiver#writeBlock doesn't throttler.
That PIPELINE_SETUP_APPEND_RECOVERY or PIPELINE_SETUP_STREAMING_RECOVERY
belongs to write block stage, so I think this stage need not throttle.
If add throttler in stage of PIPELINE_SETUP_APPEND_RECOVERY or
PIPELINE_SETUP_STREAMING_RECOVERY , I think blockReceiver.receiveBlock also
modifies and we should do it on another jira.
Please correct me if I was wrong. Thank you.
> Add DataTransferThrottler to the Datanode transfers
> ---------------------------------------------------
>
> Key: HDFS-12904
> URL: https://issues.apache.org/jira/browse/HDFS-12904
> Project: Hadoop HDFS
> Issue Type: New Feature
> Components: datanode
> Reporter: Íñigo Goiri
> Assignee: Lisheng Sun
> Priority: Minor
> Attachments: HDFS-12904.000.patch, HDFS-12904.001.patch,
> HDFS-12904.002.patch, HDFS-12904.003.patch, HDFS-12904.005.patch,
> HDFS-12904.006.patch
>
>
> The {{DataXceiverServer}} already uses throttling for the balancing. The
> Datanode should also allow throttling the regular data transfers.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
