Hadoop HDFS
  1. Hadoop HDFS
  2. HDFS-3175

When the disk space is available back,Namenode resource monitor can automatically take off safemode.

    Details

    • Type: Improvement Improvement
    • Status: Open
    • Priority: Major Major
    • Resolution: Unresolved
    • Affects Version/s: 0.24.0, 2.0.0-alpha
    • Fix Version/s: None
    • Component/s: namenode
    • Labels:
      None
    1. HDFS-3175.patch
      2 kB
      liaowenrui
    2. testcase
      2 kB
      liaowenrui
    3. HDFS-3175.patch
      2 kB
      liaowenrui
    4. HDFS-3175.patch
      2 kB
      liaowenrui

      Activity

      Hide
      Aaron T. Myers added a comment -

      FWIW, when the NameNodeResourceChecker was originally implemented it was a deliberate choice not to automatically exit SM if disk space late became available. Obviously we could change that now if we wanted, but I just figured I'd point it out.

      Show
      Aaron T. Myers added a comment - FWIW, when the NameNodeResourceChecker was originally implemented it was a deliberate choice not to automatically exit SM if disk space late became available. Obviously we could change that now if we wanted, but I just figured I'd point it out.
      Hide
      liaowenrui added a comment -

      Thank you very much for your comments.

      why do it was a deliberate choice not to automatically exit SM if disk space late became available?

      The NameNodeResourceChecker only check whether the disk is available or not, whether the system is automatically exit safemode or not is determined by NameNodeResourceMonitor.

      Show
      liaowenrui added a comment - Thank you very much for your comments. why do it was a deliberate choice not to automatically exit SM if disk space late became available? The NameNodeResourceChecker only check whether the disk is available or not, whether the system is automatically exit safemode or not is determined by NameNodeResourceMonitor.
      Hide
      Aaron T. Myers added a comment -

      Take a look at the end of this comment, in which Eli articulates the reasoning why not to automatically leave safe mode upon having enough disk space.

      The NameNodeResourceChecker only check whether the disk is available or not, whether the system is automatically exit safemode or not is determined by NameNodeResourceMonitor.

      You're right. My mistake.

      Show
      Aaron T. Myers added a comment - Take a look at the end of this comment , in which Eli articulates the reasoning why not to automatically leave safe mode upon having enough disk space. The NameNodeResourceChecker only check whether the disk is available or not, whether the system is automatically exit safemode or not is determined by NameNodeResourceMonitor. You're right. My mistake.
      Hide
      liaowenrui added a comment -

      thank you very much

      Show
      liaowenrui added a comment - thank you very much
      Hide
      liaowenrui added a comment -

      modify format

      Show
      liaowenrui added a comment - modify format
      Hide
      liaowenrui added a comment -

      hi Aaron,

      I think that automatically leave safemode mechanism is better than manual.and It should be optimized to automatical approach.

      Is it right?

      Show
      liaowenrui added a comment - hi Aaron, I think that automatically leave safemode mechanism is better than manual.and It should be optimized to automatical approach. Is it right?
      Show
      liaowenrui added a comment - Hi Aaron, I have seen your opinion in other issue HDFS-2914 https://issues.apache.org/jira/browse/HDFS-2914?focusedCommentId=13203135&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13203135 This can solve that issue as well.

        People

        • Assignee:
          Unassigned
          Reporter:
          liaowenrui
        • Votes:
          0 Vote for this issue
          Watchers:
          6 Start watching this issue

          Dates

          • Created:
            Updated:

            Development