Flume
  1. Flume
  2. FLUME-481

Unmapping a logic node leaves node in node status list

    Details

    • Type: Bug Bug
    • Status: Reopened
    • Priority: Minor Minor
    • Resolution: Unresolved
    • Affects Version/s: v0.9.4
    • Fix Version/s: v0.9.5
    • Component/s: Master
    • Labels:
      None

      Description

      If you unmap a logic node it remains in the master node status list until the master is restarted. Without looking at code I'd imagine this may be partially related to FLUME-108/FLUME-150.

      I think it should eather remove right away, or after a delayed time. However it should always remove from 1 master if it shows up on another.

        Activity

        Hide
        Jonathan Hsieh added a comment -

        There is the 'purge' command implemented in FLUME-240, that removes all the node statuses and waits for a refresh.

        Is this sufficient?

        Show
        Jonathan Hsieh added a comment - There is the 'purge' command implemented in FLUME-240 , that removes all the node statuses and waits for a refresh. Is this sufficient?
        Hide
        Nicholas Verbeck added a comment -

        That may work for now. Does this broadcast to all masters or just the connected master? What is the idea behind keeping these noncommissioned node around in the status list anyway? As well as could we not extend the unmap command to clear a node from the status list when ran? Since most likely when you remap it it will appear on a different master.

        Show
        Nicholas Verbeck added a comment - That may work for now. Does this broadcast to all masters or just the connected master? What is the idea behind keeping these noncommissioned node around in the status list anyway? As well as could we not extend the unmap command to clear a node from the status list when ran? Since most likely when you remap it it will appear on a different master.
        Hide
        Jonathan Hsieh added a comment -

        It does not broadcast to all masters. I'm currently considering a new version of the master that properly supports multi-master completely across all aspects, where that problem would get specifically addressed.

        Show
        Jonathan Hsieh added a comment - It does not broadcast to all masters. I'm currently considering a new version of the master that properly supports multi-master completely across all aspects, where that problem would get specifically addressed.
        Hide
        Steve Hoffman added a comment - - edited

        I have found that the purge command removes the node from the first table (node status), but the entry in the 3rd table remains (Physical/Logical Node mapping).
        I have not found a way to remove an old node from this list even though it has been unconfig/unmap/decommission/and purged.

        I don't know if it matters that the physical & logical names are the same, but they are in my case if that helps. Also still on CDH3u0.

        Show
        Steve Hoffman added a comment - - edited I have found that the purge command removes the node from the first table (node status), but the entry in the 3rd table remains (Physical/Logical Node mapping). I have not found a way to remove an old node from this list even though it has been unconfig/unmap/decommission/and purged. I don't know if it matters that the physical & logical names are the same, but they are in my case if that helps. Also still on CDH3u0.
        Hide
        Steve Hoffman added a comment -

        After 7 months of being open, can we get a little context around the "won't fix"? Is this because of flume-ng and flume original's multi-master being abandoned?

        Show
        Steve Hoffman added a comment - After 7 months of being open, can we get a little context around the "won't fix"? Is this because of flume-ng and flume original's multi-master being abandoned?
        Hide
        Arvind Prabhakar added a comment -

        @Steve - I think this is a mistake. I am looking into further details regarding this and other Jiras that got closed and will get back soon on it.

        Show
        Arvind Prabhakar added a comment - @Steve - I think this is a mistake. I am looking into further details regarding this and other Jiras that got closed and will get back soon on it.

          People

          • Assignee:
            Unassigned
            Reporter:
            Nicholas Verbeck
          • Votes:
            0 Vote for this issue
            Watchers:
            0 Start watching this issue

            Dates

            • Created:
              Updated:

              Development