[
https://issues.apache.org/jira/browse/HDFS-17397?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17830417#comment-17830417
]
ASF GitHub Bot commented on HDFS-17397:
---------------------------------------
Hexiaoqiao commented on PR #6591:
URL: https://github.com/apache/hadoop/pull/6591#issuecomment-2017607444
@xleoken Thanks for your proposal. I am not sure this is the proper solution
for your case as @ZanderXu mentioned. IIUC, you expect to fast fail when meet
network issue between client and the first DataNode while write data to
pipeline, right? IMO, it is difficult to determine to do that because,
a. Sometimes we could not determine that it is client to the first DataNode
network issue only using the time cost of ACK. Timeout between DataNodes in
pipeline could also lead client wait time out IMO.
b. If fast fail and recovery pipeline, the time cost could be more
considerable, such as re-create pipeline and transfer data will involve more
time cost when have writen out more than 10MB.
For this case, we have discussed times, I think we need to split it to two
step, report metrics back to client, then improve strategy (fast fail or switch
dn or some other way based on different metrics). FYI.
Thanks again.
> Choose another DN as soon as possible, when encountering network issues
> -----------------------------------------------------------------------
>
> Key: HDFS-17397
> URL: https://issues.apache.org/jira/browse/HDFS-17397
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: xleoken
> Priority: Minor
> Labels: pull-request-available
> Attachments: hadoop.png
>
>
> Choose another DN as soon as possible, when encountering network issues.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
