[
https://issues.apache.org/jira/browse/HDFS-7661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15034673#comment-15034673
]
Tsz Wo Nicholas Sze commented on HDFS-7661:
-------------------------------------------
Hi Gao, the diagram looks good. It seems that the diagram does not describe
how to handle partial cell group.
For example, suppose the user calls hflush at byte position
- p0 = 64k * 7 + 10.
We know that the first 64k * 6 bytes are written to the first cell group and we
can generate parities. But the the remaining 64k + 10 bytes, should we
generate the parities for the second cell group? If we choose not to, reader
won't be able to read data if one or more data blocks are unavailable. If we
choose to generate the parities, what should we do if the user call hflush at
- p1 = 64k * 7 + 20?
We need to regenerate the parities for the second cell group and overwrite
them. We need more detail on these issues.
> 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)
