[
https://issues.apache.org/jira/browse/HDFS-6970?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14139563#comment-14139563
]
Charles Lamb commented on HDFS-6970:
------------------------------------
Oh, you're right about buffersize. I'm so used to keying in on the magenta for
unused that I failed to notice that it wasn't a change. So, yes, I agree it
shouldn't be addressed.
The whitespace I was thinking of is at line 2482, but in retrospect it improves
readability, so NM.
Thanks for the clarification on RetryStartFileException. That makes sense.
In terms of testStartFileRetry, I'm optimistic. I used to be able to reproduce
the hang easily. Now I can't.
Anyway, I like this client-side approach a lot better so thanks for working on
this.
+1, non-binding.
> Move startFile EDEK retries to the DFSClient
> --------------------------------------------
>
> Key: HDFS-6970
> URL: https://issues.apache.org/jira/browse/HDFS-6970
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: encryption
> Affects Versions: 2.5.0
> Reporter: Andrew Wang
> Assignee: Andrew Wang
> Attachments: hdfs-6970.001.patch, hdfs-6970.002.patch
>
>
> [~sureshms] pointed out that holding on to an RPC handler while talking to
> the KMS is bad, since it can exhaust the available handlers. Let's avoid this
> by doing retries at the DFSClient rather than in the RPC handler, and moving
> EDEK fetching to the background.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
