[
https://issues.apache.org/jira/browse/HDFS-13281?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16419441#comment-16419441
]
Xiao Chen commented on HDFS-13281:
----------------------------------
{quote}
{quote}in which case there will be no step #3
{quote}
Didn't get you. can you please elaborate.
{quote}
I was saying: This patch works for an hdfs client, writing to /.reserved/raw.
In this case, the client writes raw data, so no encrypt in step #3, just raw
bytes written:
Let's step back from webhdfs, because here we're changing the NN code. Say
you're using hdfs client, and you write to /.reserved/raw. Before this patch
you'll always get a feinfo and encrypt (which probably isn't correct. But from
NN's view there's no 'write-to-raw', it just resolves to a regular path). After
this patch, you get no feinfo, and just write raw (presumably encrypted) bytes
to DN. You'll have to setxattr on that file to NN, otherwise we'll end up with
a file in HDFS that's raw and doesn't have feinfo, which is essentially corrupt
data, right?
Along this line, for webhdfs/hdfs client, after the file is created and been
written (1 block closed, or file closed), and until setxattr, the file will be
readable but undecryptable.
I think at the minimum we should setxattr immediately after the file is
created. One atomic way is perhaps pass in xattr at file creation time. What do
you think [~shahrs87] [~daryn]?
> Namenode#createFile should be /.reserved/raw/ aware.
> ----------------------------------------------------
>
> Key: HDFS-13281
> URL: https://issues.apache.org/jira/browse/HDFS-13281
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: encryption
> Affects Versions: 2.8.3
> Reporter: Rushabh S Shah
> Assignee: Rushabh S Shah
> Priority: Critical
> Attachments: HDFS-13281.001.patch, HDFS-13281.002.patch
>
>
> If I want to write to /.reserved/raw/<dir> and if that directory happens to
> be in EZ, then namenode *should not* create edek and just copy the raw bytes
> from the source.
> Namenode#startFileInt should be /.reserved/raw/ aware.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
