[
https://issues.apache.org/jira/browse/HDFS-7199?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14188861#comment-14188861
]
Colin Patrick McCabe commented on HDFS-7199:
--------------------------------------------
I'm having trouble understanding this patch. Won't the exception you are
setting with {{setLastException(new IOException("DataStreamer Exception:
",e))}} overwrite the exception set on these previous lines:
{code}
if (e instanceof IOException) {
setLastException((IOException)e);
}
{code}
Wouldn't it make more sense to simply add an else statement here where we wrap
the non-IOE in an IOE?
bq. working in progress patch. I will work on creating the test case. It is a
littble bit hard.
It looks like this will end up being a 1 or 2 line patch. So we could
potentially commit this JIRA and file a follow-up JIRA for the test case. I
think it should be possible to write a good test case using Mockito or perhaps
one of the fault injectors.
> DFSOutputStream can silently drop data if DataStreamer crashes with a non-I/O
> exception
> ---------------------------------------------------------------------------------------
>
> Key: HDFS-7199
> URL: https://issues.apache.org/jira/browse/HDFS-7199
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: hdfs-client
> Affects Versions: 2.5.0
> Reporter: Jason Lowe
> Assignee: Rushabh S Shah
> Priority: Critical
> Attachments: HDFS-7199-WIP.patch
>
>
> If the DataStreamer thread encounters a non-I/O exception then it closes the
> output stream but does not set lastException. When the client later calls
> close on the output stream then it will see the stream is already closed with
> lastException == null, mistakently think this is a redundant close call, and
> fail to report any error to the client.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
