Details

    • Bug
    • Resolution: Fixed
    • Major
    • None
    • Community Edition
    • OS
    • CentOS Linux release 7.6.1810 (Core)

    Description

      Since uppdating a few Linux systems, Observium available memory graphs looks funky.
      I have attached a graph showing the problem.

       

      Attachments

        1. discovery.txt
          17 kB
        2. fITaMU4xnxdKf7AY.png
          fITaMU4xnxdKf7AY.png
          74 kB
        3. GH3f1q8FkJGjY7xz.png
          GH3f1q8FkJGjY7xz.png
          74 kB
        4. ll5to0yEufH4D90n.png
          ll5to0yEufH4D90n.png
          94 kB
        5. Screenshot 2019-09-03 at 16.02.04.png
          Screenshot 2019-09-03 at 16.02.04.png
          68 kB

        Activity

          [OBS-3090] Available memory is miscalculated

          Workaround reverted in r10544.

          landy Mike Stupalov added a comment - Workaround reverted in r10544.

          Please find attached

          discovery.txt

          cu-olly Olorin Henderson added a comment - Please find attached discovery.txt

          please attach additional debug (I need to see UCD-SNMP-MIB::version*):

          ./discovery.php -d -m mempools -h <device>

          landy Mike Stupalov added a comment - please attach additional debug (I need to see UCD-SNMP-MIB::version*): ./discovery.php -d -m mempools -h <device>

          I fully agree with Olorin's comment above. Our CentOS 7 systems updated this week started showing funky memory graphs. Observium version 20.4.10371.

          All was fine with net-snmp-1:5.7.2-43. Upgrading to net-snmp-1:5.7.2-48 caused the issue.

          michal Michal Dobrzyniecki added a comment - I fully agree with Olorin's comment above. Our CentOS 7 systems updated this week started showing funky memory graphs. Observium version 20.4.10371. All was fine with net-snmp-1:5.7.2-43. Upgrading to net-snmp-1:5.7.2-48 caused the issue.

          RHEL/CentOS have now reversed this behaviour, so the graphs again aren't calculated properly. Can we undo this patch?

          cu-olly Olorin Henderson added a comment - RHEL/CentOS have now reversed this behaviour, so the graphs again aren't calculated properly. Can we undo this patch?

          Note that this was caused by RedHat being idiots and pushing a major behaviour change in a minor patch revision that doesn't even exist in upstream net-snmp yet.

          So much for "stable api".

          adama Adam Armstrong added a comment - Note that this was caused by RedHat being idiots and pushing a major behaviour change in a minor patch revision that doesn't even exist in upstream net-snmp yet. So much for "stable api".

          Ok, should be fixed in r10024.

          Required device(s) rediscovery.

          landy Mike Stupalov added a comment - Ok, should be fixed in r10024. Required device(s) rediscovery.

          I confirm the same issue. It is present with CloudLinux 7.7, RHEL 7.7, RHEL 8.0 and soon-released CentOS 7.7 too. I haven't had it happen to CentOS 7.6 yet but that's very worrying to see it's affected too. 

          cu-olly Olorin Henderson added a comment - I confirm the same issue. It is present with CloudLinux 7.7, RHEL 7.7, RHEL 8.0 and soon-released CentOS 7.7 too. I haven't had it happen to CentOS 7.6 yet but that's very worrying to see it's affected too. 

          People

            landy Mike Stupalov
            peter.jansson@gmail.com Peter Jansson
            Votes:
            1 Vote for this issue
            Watchers:
            5 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: