[
https://issues.apache.org/jira/browse/HADOOP-16767?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17020383#comment-17020383
]
Steve Loughran commented on HADOOP-16767:
-----------------------------------------
BTW, you say "drain event". Maybe we should be specially treating "failure on
read when we are just draining the channel for efficient re-use". any error
should just be converted to a "break the connection"
> S3AInputStream reopening does not handle non IO exceptions properly
> -------------------------------------------------------------------
>
> Key: HADOOP-16767
> URL: https://issues.apache.org/jira/browse/HADOOP-16767
> Project: Hadoop Common
> Issue Type: Sub-task
> Affects Versions: 3.1.1
> Reporter: Sergei Poganshev
> Assignee: Sergei Poganshev
> Priority: Major
>
> Since only {{IOException}} is getting caught in
> [closeStream|https://github.com/apache/hadoop/blob/24080666e5e2214d4a362c889cd9aa617be5de81/hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3AInputStream.java#L579],
> {{reopen()}} fails in case {{SdkClientException}} occurs on a drain attempt.
> This leads to multiple failing retries of {{invoker.retry("read",...)}} and
> then the whole {{read()}} operation failing.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
