You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-dev@hadoop.apache.org by "David Phillips (JIRA)" <ji...@apache.org> on 2008/11/07 18:36:49 UTC

[jira] Updated: (HADOOP-4377) Race condition creating S3 buffer directory for NativeS3FileSystem

     [ https://issues.apache.org/jira/browse/HADOOP-4377?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

David Phillips updated HADOOP-4377:
-----------------------------------

    Attachment: hadoop-s3n-tmpdir.patch

Chris, you are right, they should be equivalent.  I verified that they both only do a single stat() if the directory exists.  Patch updated.

> Race condition creating S3 buffer directory for NativeS3FileSystem
> ------------------------------------------------------------------
>
>                 Key: HADOOP-4377
>                 URL: https://issues.apache.org/jira/browse/HADOOP-4377
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: fs/s3
>    Affects Versions: 0.18.1
>            Reporter: David Phillips
>            Assignee: David Phillips
>         Attachments: hadoop-s3n-tmpdir.patch, hadoop-s3n-tmpdir.patch
>
>
> The buffer directory is checked for existence, then created if it doesn't exist.  But the create can fail if the another process creates it in between.  We can fix this by checking for existence again if the create fails.  I've seen "Cannot create S3 buffer directory" occur, and this race is the most plausible explanation.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.