[
https://issues.apache.org/jira/browse/HDFS-9574?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15089536#comment-15089536
]
Kihwal Lee edited comment on HDFS-9574 at 1/8/16 5:11 PM:
----------------------------------------------------------
bq. DFSInputStream: Instead of tracking the retryList separately, would it be
easier to just add it back to the nodeList and set the isRetry boolean?
I thought about doing that. But with one list, it is hard to tell whether it
has any good candidate left to try or it has only retriable nodes. The code
ended up being more complicated than necessary so I settled with two separate
lists.
bq. DataXceiver: Might consider changing checkAccess to not require the stream
and just have it call getBufferedOutputStream.
It looks like that might be okay for the current usages. I didn't try to
improve all existing inconsistencies in there. It might be better to be done in
a separate clean-up jira.
was (Author: kihwal):
bq. DFSInputStream: Instead of tracking the retryList separately, would it be
easier to just add it back to the nodeList and set the isRetry boolean?
I thought about doing that. But with one list, it is hard to tell whether it
has any good candidate left to try or it has only retriable nodes. The code
ended up being more complicated than necessary so I settled with two separate
lists.
bq. DataXceiver: Might consider changing checkAccess to not require the stream
and just have it call getBufferedOutputStream.
It looks like that might be okay for the current usages. I didn't try to
improve all inconsistencies in there. It might be better to be done in a
separate clean-up jira.
> Reduce client failures during datanode restart
> ----------------------------------------------
>
> Key: HDFS-9574
> URL: https://issues.apache.org/jira/browse/HDFS-9574
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Kihwal Lee
> Assignee: Kihwal Lee
> Attachments: HDFS-9574.patch, HDFS-9574.v2.patch, HDFS-9574.v3.patch
>
>
> Since DataXceiverServer is initialized before BP is fully up, client requests
> will fail until the datanode registers.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
