Uploaded image for project: 'Hadoop HDFS'
  1. Hadoop HDFS
  2. HDFS-6581 Write to single replica in memory
  3. HDFS-7291

Persist in-memory replicas with appropriate unbuffered copy API on POSIX and Windows

VotersWatch issueWatchersLinkCloneUpdate Comment AuthorReplace String in CommentUpdate Comment VisibilityDelete Comments
    XMLWordPrintableJSON

Details

    • Sub-task
    • Status: Closed
    • Major
    • Resolution: Fixed
    • 2.6.0
    • 2.6.0
    • datanode
    • None
    • Reviewed

    Description

      HDFS-7090 changes to persist in-memory replicas using unbuffered IO on Linux and Windows. On Linux distribution, it relies on the sendfile() API between two file descriptors to achieve unbuffered IO copy. According to Linux document at http://man7.org/linux/man-pages/man2/sendfile.2.html, this is only supported on Linux kernel 2.6.33+.

      As pointed by Haowei in the discussion below, FileChannel#transferTo already has support for native unbuffered IO on POSIX platform. On Windows, JDK 6/7/8 has not implemented native unbuffered IO yet. We change to use FileChannel#transfer for POSIX and our own native wrapper of CopyFileEx on Windows for unbuffered copy.

      Attachments

        1. HDFS-7291.4.patch
          10 kB
          Xiaoyu Yao
        2. HDFS-7291.3.patch
          10 kB
          Xiaoyu Yao
        3. HDFS-7291.2.patch
          8 kB
          Xiaoyu Yao
        4. HDFS-7291.1.patch
          6 kB
          Xiaoyu Yao
        5. HDFS-7291.0.patch
          6 kB
          Xiaoyu Yao

        Issue Links

        Activity

          This comment will be Viewable by All Users Viewable by All Users
          Cancel

          People

            xyao Xiaoyu Yao
            xyao Xiaoyu Yao
            Votes:
            0 Vote for this issue
            Watchers:
            6 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Slack

                Issue deployment