Spikes on the graph

[petr5250]

Hi,
I'm trying to monitor MikroTik router interface statistics through SNMPv1 64bit counters (ifHCInOctets and ifHCOutOctets). For the most of the time SNMP values are retrieved correctly (average throughput on the interface is 30-40Mbps), but from time to time (which depends on the actual throughput ) there is a spike on the graph:


At first I thought that there is some problem with SNMP, but according to series of manual snmpwalks counters on the router are adjusted correctly. In the time when there is a spike on the graph, there isn't any counter overflow or other unexpected value of the SNMP counter.

Here is the definition of the data item:


I'm using Zabbix v.1.8.2 with pgsql backend and housekeeping interval of 24 hours.

Could you help me?
Thanks.

P.S.: I noticed, that the same spikes are on the predefined interface graphs of the Zabbix server (RHEL 4), which is monitored by Zabbix agent.

[Aly]

Check incoming data in latest data screen by switching to values instead of graphs in top-right corner.

[richlv]

as aly noted, check actual data.

is that host virtualised by any chance, and maybe running scheduled ntpdate instead of ntpd ? really flaky time on zabbix server could cause something similar.

ps. love the host name