[
https://issues.apache.org/jira/browse/HDFS-7661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15033589#comment-15033589
]
GAO Rui commented on HDFS-7661:
-------------------------------
Thank you very much, [~szetszwo].
I have made a draft flush/sync steps plan for EC files
[^EC-file-flush-and-sync-steps-plan-2015-12-01.png].
The plan is based on the ideas of {{DFSStripedOutputStream#closeImpl( )}} and
{{DFSOutputStream#flushOrSync( )}}. The steps in yellow ovals is the planned
EC file flush/sync steps.
The most different part between flush/sync steps with
{{DFSStripedOutputStream#closeImpl( )}} is in *step 1.2* and *step 3*.
In *step 1.2*, we plan to send an empty "signal" packet to all DN.
In *step 3*, we do not close streamers. And we update visible length if
required.
I am aware of that, besides the plan steps, there are a lot of other details
should be taken care of. Before diving into the implementing codes in detail. I
want to confirm the big picture of the process with you. Please feel free to
comment and conduct any concerns you have.
Thank you very much.
> Support read when a EC file is being written
> --------------------------------------------
>
> Key: HDFS-7661
> URL: https://issues.apache.org/jira/browse/HDFS-7661
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Tsz Wo Nicholas Sze
> Assignee: GAO Rui
> Attachments: EC-file-flush-and-sync-steps-plan-2015-12-01.png,
> HDFS-7661-unitTest-wip-trunk.patch
>
>
> We also need to support hflush/hsync and visible length.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
