Ad Widget

Collapse

missing (graph) data

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • erikklaassen
    Junior Member
    • Sep 2007
    • 13

    #1

    missing (graph) data

    Hi,

    I have a zabbix server witch is monitoring 35 switches, a total of 1300 items. We monitor al our switches with snmp and take samples once every 5 min.

    On interfaces that have more than 100mbit/sec the graphs starting to show gaps. It is actually not a graph problem because the latest data is also showing no data:

    H_swithX: Port 169 GigabitEthernet3/2 OutOctets

    2011-10-26 12:28:08 1319624888 92293392
    2011-10-26 12:03:08 1319623388 101673328
    2011-10-26 11:58:08 1319623088 1905976
    2011-10-26 11:53:08 1319622788 2304328
    2011-10-26 11:43:08 1319622188 68140544

    It should have data every 300sec.

    I updated the StartPollers to 50, no luck...
    I checked the queue's most of the time they are empty, sometimes snmpv2 shows ~30 items in the 5 sec queue.

    The zabbixserver(1.8.6) is a VM on a XenServer, using 2 core's and 4gb memory.
    Last edited by erikklaassen; 01-11-2011, 11:04.
  • jobster
    Junior Member
    • Sep 2008
    • 4

    #2
    i have a similar setup, with the same problem, this most often seen on our 10GE uplinks and on etherchannels where there is more traffic

    Number of hosts (monitored/not monitored/templates) 652 605 / 9 / 38
    Number of items (monitored/disabled/not supported) 201948 161320 / 39495 / 1133
    Number of triggers (enabled/disabled)[problem/unknown/ok] 2295 1764 / 531 [2 / 117 / 1645]
    Number of users (online) 14 2
    Required server performance, new values per second 302.64 -

    Comment

    • erikklaassen
      Junior Member
      • Sep 2007
      • 13

      #3
      Yes it seem only on the interfaces that handle more than 100Mbit/sec. I checked other interfaces during the same period of time and they are not missing any data.
      When the traffic on the 1Ge interface drops from 150Mb/sec to about 50 the data is collected fine.

      Has i something to do with the large number of octets and the custom multiplier? I collect the octets and multiply them by 8 and store them as delta speed per second.

      Comment

      • erikklaassen
        Junior Member
        • Sep 2007
        • 13

        #4
        Found it!

        This document describes answers to commonly asked questions about SNMP counters as they relate to Cisco equipment.


        It seems when speed goes over 100Mb/s the octect counters wrap faster than 300 second. I think Zabbix cant calculate this anymore when counters wrap to fast

        We have to use Highspeed Counters like IfHCinOctets. I will test this and report tomorrow

        Comment

        • erikklaassen
          Junior Member
          • Sep 2007
          • 13

          #5
          Using the high speed counters did indeed do the trick:

          Comment

          Working...