[
https://issues.apache.org/jira/browse/HDFS-7661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15035153#comment-15035153
]
GAO Rui commented on HDFS-7661:
-------------------------------
bq. The data length could be used as the flag. Parities correspond to data
length. We may store the data length along with a party for being written
blocks.
Cause the parity cells will always be full, so I guess you mean we store the
data length along with the partial cell which is in the end of file data, right?
So there is a minor possibility that the partial data cell size is almost the
full cell size. For example, 63K + 1023B+7b, so the partial cell plus the data
length will exceed the full cell size. I am not sure whether is possible to
store a cell which exceeded the full cell size. I will look at the
{{DFSStripedInputStream.StripeReader.readStripe()}} and DN cell writing code.
Thank you very much for the tips.
> 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)
