Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-4617

Couchbase Server doesn't pick up available disk space from automount directory

    Details

    • Type: Bug
    • Status: Resolved
    • Priority: Critical
    • Resolution: Fixed
    • Affects Version/s: 1.7.2, 1.8.1, 2.0, 2.0.1, 2.1.0, 2.2.0, 2.5.0
    • Fix Version/s: 3.0
    • Component/s: ns_server
    • Security Level: Public
    • Labels:
    • Triage:
      Untriaged
    • Is this a Regression?:
      Yes
    • Sprint:
      Sprint 2 - March 11 - April 3

      Description

      /home/fds/membase is an automount entry that mounts the /export filesystem (specifically, /export/fds/Linux_RHEL5_x86_64/membase). 'df' does show the proper location, although the output varies depending on the path that is passed as an argument.

      ]$ cd /home/fds/membase

      $ df -h .
      Filesystem Size Used Avail Use% Mounted on
      /dev/mapper/vg1-lvexport
      992G 33G 908G 4% /export

      $ df -h /home/fds/membase
      Filesystem Size Used Avail Use% Mounted on
      /export/fds/Linux_RHEL5_x86_64/membase
      992G 33G 908G 4% /home/fds/membase

      Membase has been setup to use /home/fds/membase, but the console incorrectly shows 6GB which is the available space on '/'

        Issue Links

          Activity

          Hide
          alkondratenko Aleksey Kondratenko (Inactive) added a comment -

          I understand that some people would like this fixed. And fix is relatively simple I.e. just fork erlang's disk_sup and pass right flags to du.

          But I'd like as much as possible to avoid doing that. Because that would mean we would have to track any upstream changes and integrate them with our fork.

          Now why this is not a blocker.

          Problem only happens on older versions of coreutils (sadly rhel6 is not new enough but I bet rhel 7 will be new enough). And problem only happens on nfs mounts and on bind mounts.

          We officially support neither.

          Hopefully things are more clear now.

          Show
          alkondratenko Aleksey Kondratenko (Inactive) added a comment - I understand that some people would like this fixed. And fix is relatively simple I.e. just fork erlang's disk_sup and pass right flags to du. But I'd like as much as possible to avoid doing that. Because that would mean we would have to track any upstream changes and integrate them with our fork. Now why this is not a blocker. Problem only happens on older versions of coreutils (sadly rhel6 is not new enough but I bet rhel 7 will be new enough). And problem only happens on nfs mounts and on bind mounts. We officially support neither. Hopefully things are more clear now.
          Hide
          dipti Dipti Borkar added a comment -

          Re-consider for next minor or major release.

          Show
          dipti Dipti Borkar added a comment - Re-consider for next minor or major release.
          Hide
          brent Brent Woodruff (Inactive) added a comment -

          Question on this issue: Could we avoid needing to fork erlang's disk_sup if we allowed overriding the device that should be queried for data and/or view disk space? The idea would be to say "if you want to do clever things like bind mounts, you have to tell us explicitly what device you want checked for space." We are probably going to support and run on RHEL 5 and 6 (and their old coreutils) for a long time.

          Show
          brent Brent Woodruff (Inactive) added a comment - Question on this issue: Could we avoid needing to fork erlang's disk_sup if we allowed overriding the device that should be queried for data and/or view disk space? The idea would be to say "if you want to do clever things like bind mounts, you have to tell us explicitly what device you want checked for space." We are probably going to support and run on RHEL 5 and 6 (and their old coreutils) for a long time.
          Hide
          alkondratenko Aleksey Kondratenko (Inactive) added a comment -

          >> Could we avoid needing to fork erlang's disk_sup if we allowed overriding the device that should be queried for data and/or view disk space?

          Don't think it's good idea. That's extra complexity for no good reason. We'll have to maintain it, expose cli/REST-API for that and do all that "enterprisey cost" about keeping backwards compat etc. And it doesn't help with NFS or other network fses.

          Show
          alkondratenko Aleksey Kondratenko (Inactive) added a comment - >> Could we avoid needing to fork erlang's disk_sup if we allowed overriding the device that should be queried for data and/or view disk space? Don't think it's good idea. That's extra complexity for no good reason. We'll have to maintain it, expose cli/REST-API for that and do all that "enterprisey cost" about keeping backwards compat etc. And it doesn't help with NFS or other network fses.
          Show
          artem Artem Stemkovski added a comment - http://review.couchbase.org/37488

            People

            • Assignee:
              thuan Thuan Nguyen
              Reporter:
              perry Perry Krug
            • Votes:
              0 Vote for this issue
              Watchers:
              11 Start watching this issue

              Dates

              • Created:
                Updated:
                Resolved:

                Agile