Agent vs SNMP and different graphs

[overnet]

I have a problem with graphs drawn using SNMP (collected data) vs graphs drawn using agents.

SNMP graphs are scattered while agent graphs are smooth.

Please see attachment.

Any thoughts on why this the case?

Thanks

[richlv]

quite likely overloaded snmp agent that can't respond to all queries in a timely manner.

[overnet]

quite likely overloaded snmp agent that can't respond to all queries in a timely manner.

Can you please elaborate more on what do you mean by "overloaded"?

I've ran 100 query using snmpwalk and I got values every time. (scripted)

[richlv]

against the same device ?
how often do you query each item on that device from within zabbix, and how many items do you monitor ?

[nelsonab]

I'm thinking along the same lines as Rich here. There's a possibility that your server is starving for pollers. How many pollers is your system set for? Under Reports -> Status of Zabbix how many values per second are being pulled down?

[overnet]

against the same device ?
how often do you query each item on that device from within zabbix, and how many items do you monitor ?

Yes.
Some items are queried every 5 min some items every 30 sec, I currently have 17 items that I monitor using SNMP, I've added those items anything that came with zabbix by default were disabled.

Under Reports -> Status of Zabbix how many values per second are being pulled down?

19.2661

Thanks,

[neoyagami]

i got the same problem. aside i got a cacti without this problem.

[richlv]

Yes.
Some items are queried every 5 min some items every 30 sec, I currently have 17 items that I monitor using SNMP, I've added those items anything that came with zabbix by default were disabled.

ok, then probably far from overloading
anything interesting in zabbix server log ?
you could try turning debugging level up.

[Slowride]

I'd flip the 'configuration/graphs/interface/Y axis type' from 'calculated' to 'calculated [min=0] before I tried to much.

From your picts, the fact your interface hit 109Mb/s insinuates its a gig interface. My MRTG graphs capped right around there until I switched from 32 bit counters to the 64bit ifHC counters.

I'm still struggling to get Zabbix to graph a gig interface running >300Mb/s.