[
https://issues.apache.org/jira/browse/HDFS-7748?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14369817#comment-14369817
]
Chris Nauroth commented on HDFS-7748:
-------------------------------------
Hi Anu. This patch looks mostly good to me. Here are a few comments.
# {{BlockReceiver.PacketResponder#run}}: There is a line that calculates a
{{myStatus}} local variable. This had been introduced in HDFS-7270. I think
it's unused now and can be removed.
# datatransfer.proto: There is typo in the comment: "configration" instead of
"configuration". Please indent by 4 spaces. (Sorry to nitpick.)
# hdfs-site.xml: Please remove this file from the next revision of the patch.
(I know you had been doing some local testing with the ECN configuration turned
on.)
I think the original intent of HDFS-7270 was to try to match the semantics of
RFC-3168 as closely as possible. Therefore, the {{ECN}} enum needed 4 possible
values, which were named {{DISABLED}}, {{SUPPORTED}}, {{SUPPORTED2}} and
{{CONGESTED}}. Shall we keep the same values here in the new protobuf enum?
I'm not very familiar with this RFC, so I'll defer to [~wheat9] on what is the
best definition of these values.
{{TestTracing}} fails for me on current trunk even without this patch. All
other test failures appear to be unrelated.
> Separate ECN flags from the Status in the DataTransferPipelineAck
> -----------------------------------------------------------------
>
> Key: HDFS-7748
> URL: https://issues.apache.org/jira/browse/HDFS-7748
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Haohui Mai
> Assignee: Anu Engineer
> Priority: Blocker
> Attachments: hdfs-7748.001.patch
>
>
> Prior to the discussions on HDFS-7270, the old clients might fail to talk to
> the newer server when ECN is turned on. This jira proposes to separate the
> ECN flags in a separate protobuf field to make the ack compatible on both
> versions.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
